fix indent in evict.cpp

Check the validity of the value before performing the create operation,
prevents new data from being generated even if the request fails to execute.

Co-authored-by: Oran Agra <oran@redislabs.com>
Co-authored-by: chendianqiang <chendianqiang@meituan.com>
Co-authored-by: Binbin <binloveplay1314@qq.com>

deps/Makefile

@@ -95,7 +95,7 @@ JEMALLOC_LDFLAGS= $(LDFLAGS)
 
 jemalloc: .make-prerequisites
 	@printf '%b %b\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)
-	cd jemalloc && ./configure --with-version=5.2.1-0-g0 --with-lg-quantum=3 --with-jemalloc-prefix=je_ --disable-cxx CFLAGS="$(JEMALLOC_CFLAGS)" LDFLAGS="$(JEMALLOC_LDFLAGS)"
+	cd jemalloc && ./configure --with-version=5.2.1-0-g0 --with-lg-quantum=3 --disable-cxx CFLAGS="$(JEMALLOC_CFLAGS)" LDFLAGS="$(JEMALLOC_LDFLAGS)"
 	cd jemalloc && $(MAKE) CFLAGS="$(JEMALLOC_CFLAGS)" LDFLAGS="$(JEMALLOC_LDFLAGS)" lib/libjemalloc.a
 
 .PHONY: jemalloc

runtest-moduleapi

@@ -37,5 +37,6 @@ $TCLSH tests/test_helper.tcl \
 --single unit/moduleapi/hash \
 --single unit/moduleapi/zset \
 --single unit/moduleapi/stream \
+--single unit/moduleapi/load \
 --config server-threads 3 \
 "${@}"

src/IStorage.h

@@ -31,6 +31,7 @@ public:
     virtual void retrieve(const char *key, size_t cchKey, callbackSingle fn) const = 0;
     virtual size_t clear() = 0;
     virtual bool enumerate(callback fn) const = 0;
+    virtual size_t stateful_enumerate(callback fn) = 0;
     virtual size_t count() const = 0;
 
     virtual void bulkInsert(char **rgkeys, size_t *rgcbkeys, char **rgvals, size_t *rgcbvals, size_t celem) {

src/Makefile

@@ -25,7 +25,7 @@ NODEPS:=clean distclean
 
 # Default settings
 STD=-pedantic -DREDIS_STATIC=''
-CXX_STD=-std=c++14 -pedantic -fno-rtti -D__STDC_FORMAT_MACROS  
+CXX_STD=-std=c++17 -pedantic -fno-rtti -D__STDC_FORMAT_MACROS  
 ifneq (,$(findstring clang,$(CC)))
   STD+=-Wno-c11-extensions
 else

src/SnapshotPayloadParseState.cpp

@@ -140,7 +140,7 @@ void SnapshotPayloadParseState::flushQueuedKeys() {
     auto &insertsInFlightTmp = insertsInFlight; // C++ GRRRRRRRRRRRRRRRR, we don't want to capute "this" because that's dangerous
     if (current_database < cserver.dbnum) {
         g_pserver->asyncworkqueue->AddWorkFunction([idb, vecqueuedKeys = std::move(this->vecqueuedKeys), vecqueuedKeysCb = std::move(this->vecqueuedKeysCb), vecqueuedVals = std::move(this->vecqueuedVals), vecqueuedValsCb = std::move(this->vecqueuedValsCb), &insertsInFlightTmp, pallocator = m_spallocator.release()]() mutable {
-            g_pserver->db[idb]->bulkStorageInsert(vecqueuedKeys.data(), vecqueuedKeysCb.data(), vecqueuedVals.data(), vecqueuedValsCb.data(), vecqueuedKeys.size());
+            g_pserver->db[idb]->bulkDirectStorageInsert(vecqueuedKeys.data(), vecqueuedKeysCb.data(), vecqueuedVals.data(), vecqueuedValsCb.data(), vecqueuedKeys.size());
             --insertsInFlightTmp;
             delete pallocator;
         });

src/StorageCache.h

@@ -49,6 +49,7 @@ public:
     void expand(uint64_t slots);
 
     bool enumerate(IStorage::callback fn) const { return m_spstorage->enumerate(fn); }
+    size_t stateful_enumerate(IStorage::callback fn) { return m_spstorage->stateful_enumerate(fn); }
 
     void beginWriteBatch();
     void endWriteBatch() { m_spstorage->endWriteBatch(); }

src/ae.cpp

@@ -716,11 +716,6 @@ int aeProcessEvents(aeEventLoop *eventLoop, int flags)
             }
         }
 
-        if (eventLoop->flags & AE_DONT_WAIT) {
-            tv.tv_sec = tv.tv_usec = 0;
-            tvp = &tv;
-        }
-
         if (eventLoop->beforesleep != NULL && flags & AE_CALL_BEFORE_SLEEP) {
             std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock);
             if (!(eventLoop->beforesleepFlags & AE_SLEEP_THREADSAFE)) {
@@ -731,6 +726,11 @@ int aeProcessEvents(aeEventLoop *eventLoop, int flags)
             eventLoop->beforesleep(eventLoop);
         }
 
+        if (eventLoop->flags & AE_DONT_WAIT) {
+            tv.tv_sec = tv.tv_usec = 0;
+            tvp = &tv;
+        }
+
         /* Call the multiplexing API, will return only on timeout or when
          * some event fires. */
         numevents = aeApiPoll(eventLoop, tvp);
@@ -859,6 +859,11 @@ void aeReleaseForkLock()
     g_forkLock.downgradeWrite();
 }
 
+void aeForkLockInChild()
+{
+    g_forkLock.setNotify(false);
+}
+
 int aeThreadOwnsLock()
 {
     return fOwnLockOverride || g_lock.fOwnLock();

src/ae.h

@@ -171,6 +171,7 @@ int aeTryAcquireLock(int fWeak);
 void aeThreadOffline();
 void aeReleaseLock();
 void aeReleaseForkLock();
+void aeForkLockInChild();
 int aeThreadOwnsLock();
 void aeSetThreadOwnsLockOverride(int fOverride);
 int aeLockContested(int threshold);

src/anet.c

@@ -441,7 +441,7 @@ static int _anetTcpServer(char *err, int port, const char *bindaddr, int af, int
     char _port[6];  /* strlen("65535") */
     struct addrinfo hints, *servinfo, *p;
 
-    snprintf(_port,6,"%d",port);
+    snprintf(_port,sizeof(_port),"%d",port);
     memset(&hints,0,sizeof(hints));
     hints.ai_family = af;
     hints.ai_socktype = SOCK_STREAM;

src/aof.cpp

@@ -285,7 +285,7 @@ int startAppendOnly(void) {
             strerror(errno));
         return C_ERR;
     }
-    if (hasActiveChildProcess() && g_pserver->child_type != CHILD_TYPE_AOF) {
+    if (hasActiveChildProcessOrBGSave() && g_pserver->child_type != CHILD_TYPE_AOF) {
         g_pserver->aof_rewrite_scheduled = 1;
         serverLog(LL_WARNING,"AOF was enabled but there is already another background operation. An AOF background was scheduled to start when possible.");
     } else {
@@ -438,7 +438,7 @@ void flushAppendOnlyFile(int force) {
      * useful for graphing / monitoring purposes. */
     if (sync_in_progress) {
         latencyAddSampleIfNeeded("aof-write-pending-fsync",latency);
-    } else if (hasActiveChildProcess()) {
+    } else if (hasActiveChildProcessOrBGSave()) {
         latencyAddSampleIfNeeded("aof-write-active-child",latency);
     } else {
         latencyAddSampleIfNeeded("aof-write-alone",latency);
@@ -535,7 +535,7 @@ void flushAppendOnlyFile(int force) {
 try_fsync:
     /* Don't fsync if no-appendfsync-on-rewrite is set to yes and there are
      * children doing I/O in the background. */
-    if (g_pserver->aof_no_fsync_on_rewrite && hasActiveChildProcess())
+    if (g_pserver->aof_no_fsync_on_rewrite && hasActiveChildProcessOrBGSave())
         return;
 
     /* Perform the fsync if needed. */
@@ -752,7 +752,7 @@ void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int a
      * accumulate the differences between the child DB and the current one
      * in a buffer, so that when the child process will do its work we
      * can append the differences to the new append only file. */
-    if (g_pserver->child_type == CHILD_TYPE_AOF)
+    if (hasActiveChildProcess() && g_pserver->child_type == CHILD_TYPE_AOF)
         aofRewriteBufferAppend((unsigned char*)buf,sdslen(buf));
 
     sdsfree(buf);
@@ -1658,7 +1658,7 @@ int rewriteAppendOnlyFile(char *filename) {
 { // BEGIN GOTO SCOPED VARIABLES
     /* Note that we have to use a different temp name here compared to the
      * one used by rewriteAppendOnlyFileBackground() function. */
-    snprintf(tmpfile,256,"temp-rewriteaof-%d.aof", (int) getpid());
+    snprintf(tmpfile,sizeof(tmpfile),"temp-rewriteaof-%d.aof", (int) getpid());
     fp = fopen(tmpfile,"w");
     if (!fp) {
         serverLog(LL_WARNING, "Opening the temp file for AOF rewrite in rewriteAppendOnlyFile(): %s", strerror(errno));
@@ -1887,7 +1887,7 @@ void aofClosePipes(void) {
 int rewriteAppendOnlyFileBackground(void) {
     pid_t childpid;
 
-    if (hasActiveChildProcess()) return C_ERR;
+    if (hasActiveChildProcessOrBGSave()) return C_ERR;
     if (aofCreatePipes() != C_OK) return C_ERR;
     if ((childpid = redisFork(CHILD_TYPE_AOF)) == 0) {
         char tmpfile[256];
@@ -1895,7 +1895,7 @@ int rewriteAppendOnlyFileBackground(void) {
         /* Child */
         redisSetProcTitle("keydb-aof-rewrite");
         redisSetCpuAffinity(g_pserver->aof_rewrite_cpulist);
-        snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) getpid());
+        snprintf(tmpfile,sizeof(tmpfile),"temp-rewriteaof-bg-%d.aof", (int) getpid());
         if (rewriteAppendOnlyFile(tmpfile) == C_OK) {
             sendChildCowInfo(CHILD_INFO_TYPE_AOF_COW_SIZE, "AOF rewrite");
             exitFromChild(0);
@@ -1930,7 +1930,7 @@ int rewriteAppendOnlyFileBackground(void) {
 void bgrewriteaofCommand(client *c) {
     if (g_pserver->child_type == CHILD_TYPE_AOF) {
         addReplyError(c,"Background append only file rewriting already in progress");
-    } else if (hasActiveChildProcess()) {
+    } else if (hasActiveChildProcessOrBGSave()) {
         g_pserver->aof_rewrite_scheduled = 1;
         addReplyStatus(c,"Background append only file rewriting scheduled");
     } else if (rewriteAppendOnlyFileBackground() == C_OK) {
@@ -1944,10 +1944,10 @@ void bgrewriteaofCommand(client *c) {
 void aofRemoveTempFile(pid_t childpid) {
     char tmpfile[256];
 
-    snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) childpid);
+    snprintf(tmpfile,sizeof(tmpfile),"temp-rewriteaof-bg-%d.aof", (int) childpid);
     bg_unlink(tmpfile);
 
-    snprintf(tmpfile,256,"temp-rewriteaof-%d.aof", (int) childpid);
+    snprintf(tmpfile,sizeof(tmpfile),"temp-rewriteaof-%d.aof", (int) childpid);
     bg_unlink(tmpfile);
 }
 
@@ -1985,7 +1985,7 @@ void backgroundRewriteDoneHandler(int exitcode, int bysignal) {
         /* Flush the differences accumulated by the parent to the
          * rewritten AOF. */
         latencyStartMonitor(latency);
-        snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof",
+        snprintf(tmpfile,sizeof(tmpfile),"temp-rewriteaof-bg-%d.aof",
             (int)g_pserver->child_pid);
         newfd = open(tmpfile,O_WRONLY|O_APPEND);
         if (newfd == -1) {

src/childinfo.cpp

@@ -42,6 +42,7 @@ typedef struct {
  * RDB / AOF saving process from the child to the parent (for instance
  * the amount of copy on write memory used) */
 void openChildInfoPipe(void) {
+    serverAssert(g_pserver->child_info_pipe[0] == -1);
     if (pipe(g_pserver->child_info_pipe) == -1) {
         /* On error our two file descriptors should be still set to -1,
          * but we call anyway closeChildInfoPipe() since can't hurt. */

src/cluster.cpp

@@ -5107,7 +5107,7 @@ void createDumpPayload(rio *payload, robj_roptr o, robj *key) {
     serverAssert(rdbSaveObject(payload,o,key));
     char szT[32];
     uint64_t mvcc = mvccFromObj(o);
-    snprintf(szT, 32, "%" PRIu64, mvcc);
+    snprintf(szT, sizeof(szT), "%" PRIu64, mvcc);
     serverAssert(rdbSaveAuxFieldStrStr(payload,"mvcc-tstamp", szT) != -1);
 
     /* Write the footer, this is how it looks like:

src/config.cpp

@@ -35,11 +35,13 @@
 
 #include <fcntl.h>
 #include <sys/stat.h>
+#include <sys/wait.h>
 #ifdef __linux__
 #include <sys/sysinfo.h>
 #endif
 
 const char *KEYDB_SET_VERSION = KEYDB_REAL_VERSION;
+size_t g_semiOrderedSetTargetBucketSize = 0;    // Its a header only class so nowhere else for this to go
 
 /*-----------------------------------------------------------------------------
  * Config file name-value maps.
@@ -2455,6 +2457,32 @@ static int isValidAOFfilename(char *val, const char **err) {
     return 1;
 }
 
+static int isValidS3Bucket(char *s3bucket, const char **err) {
+    int status = EXIT_FAILURE;
+    pid_t pid = fork();
+    if (pid < 0)
+    {
+        *err = "couldn't fork to call aws cli";
+        return 0;
+    }
+
+    if (pid == 0)
+    {
+        execlp("aws", "aws", "s3", "ls", s3bucket);
+        exit(EXIT_FAILURE);
+    }
+    else 
+    {
+        waitpid(pid, &status, 0);
+    }
+
+    if (status != EXIT_SUCCESS) {
+        *err = "could not access s3 bucket";
+        return 0;
+    }
+    return 1;
+}
+
 /* Validate specified string is a valid proc-title-template */
 static int isValidProcTitleTemplate(char *val, const char **err) {
     if (!validateProcTitleTemplate(val)) {
@@ -2587,7 +2615,7 @@ static int updateMaxclients(long long val, long long prev, const char **err) {
         adjustOpenFilesLimit();
         if (g_pserver->maxclients != val) {
             static char msg[128];
-            snprintf(msg, 128, "The operating system is not able to handle the specified number of clients, try with %d", g_pserver->maxclients);
+            snprintf(msg, sizeof(msg), "The operating system is not able to handle the specified number of clients, try with %d", g_pserver->maxclients);
             *err = msg;
             if (g_pserver->maxclients > prev) {
                 g_pserver->maxclients = prev;
@@ -2626,7 +2654,7 @@ static int updateMaxclients(long long val, long long prev, const char **err) {
 
                 if (res != AE_OK){
                     static char msg[128];
-                    snprintf(msg, 128,"Failed to post the request to change setsize for Thread %d", iel);
+                    snprintf(msg, sizeof(msg),"Failed to post the request to change setsize for Thread %d", iel);
                     *err = msg;
                     return 0;
                 }
@@ -2778,7 +2806,7 @@ standardConfig configs[] = {
     createBoolConfig("appendonly", NULL, MODIFIABLE_CONFIG, g_pserver->aof_enabled, 0, NULL, updateAppendonly),
     createBoolConfig("cluster-allow-reads-when-down", NULL, MODIFIABLE_CONFIG, g_pserver->cluster_allow_reads_when_down, 0, NULL, NULL),
     createBoolConfig("delete-on-evict", NULL, MODIFIABLE_CONFIG, cserver.delete_on_evict, 0, NULL, NULL),
-    createBoolConfig("use-fork", NULL, IMMUTABLE_CONFIG, cserver.fForkBgSave, 0, NULL, NULL),
+    createBoolConfig("use-fork", NULL, IMMUTABLE_CONFIG, cserver.fForkBgSave, 1, NULL, NULL),
     createBoolConfig("io-threads-do-reads", NULL, IMMUTABLE_CONFIG, fDummy, 0, NULL, NULL),
     createBoolConfig("time-thread-priority", NULL, IMMUTABLE_CONFIG, cserver.time_thread_priority, 0, NULL, NULL),
     createBoolConfig("prefetch-enabled", NULL, MODIFIABLE_CONFIG, g_pserver->prefetch_enabled, 1, NULL, NULL),
@@ -2802,6 +2830,7 @@ standardConfig configs[] = {
     createStringConfig("cluster-announce-ip", NULL, MODIFIABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->cluster_announce_ip, NULL, NULL, NULL),
     createStringConfig("syslog-ident", NULL, IMMUTABLE_CONFIG, ALLOW_EMPTY_STRING, g_pserver->syslog_ident, "redis", NULL, NULL),
     createStringConfig("dbfilename", NULL, MODIFIABLE_CONFIG, ALLOW_EMPTY_STRING, g_pserver->rdb_filename, CONFIG_DEFAULT_RDB_FILENAME, isValidDBfilename, NULL),
+    createStringConfig("db-s3-object", NULL, MODIFIABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->rdb_s3bucketpath, NULL, isValidS3Bucket, NULL),
     createStringConfig("appendfilename", NULL, IMMUTABLE_CONFIG, ALLOW_EMPTY_STRING, g_pserver->aof_filename, "appendonly.aof", isValidAOFfilename, NULL),
     createStringConfig("server_cpulist", NULL, IMMUTABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->server_cpulist, NULL, NULL, NULL),
     createStringConfig("bio_cpulist", NULL, IMMUTABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->bio_cpulist, NULL, NULL, NULL),
@@ -2883,6 +2912,7 @@ standardConfig configs[] = {
     createLongLongConfig("repl-backlog-size", NULL, MODIFIABLE_CONFIG, 1, LLONG_MAX, g_pserver->repl_backlog_size, 1024*1024, MEMORY_CONFIG, NULL, updateReplBacklogSize), /* Default: 1mb */
     createLongLongConfig("repl-backlog-disk-reserve", NULL, IMMUTABLE_CONFIG, 0, LLONG_MAX, cserver.repl_backlog_disk_size, 0, MEMORY_CONFIG, NULL, NULL),
     createLongLongConfig("max-snapshot-slip", NULL, MODIFIABLE_CONFIG, 0, 5000, g_pserver->snapshot_slip, 400, 0, NULL, NULL),
+    createLongLongConfig("max-rand-count", NULL, MODIFIABLE_CONFIG, 0, LONG_MAX/2, g_pserver->rand_total_threshold, LONG_MAX/2, 0, NULL, NULL),
 
     /* Unsigned Long Long configs */
     createULongLongConfig("maxmemory", NULL, MODIFIABLE_CONFIG, 0, LLONG_MAX, g_pserver->maxmemory, 0, MEMORY_CONFIG, NULL, updateMaxmemory),
@@ -2910,6 +2940,7 @@ standardConfig configs[] = {
     createBoolConfig("allow-write-during-load", NULL, MODIFIABLE_CONFIG, g_pserver->fWriteDuringActiveLoad, 0, NULL, NULL),
     createBoolConfig("force-backlog-disk-reserve", NULL, MODIFIABLE_CONFIG, cserver.force_backlog_disk, 0, NULL, NULL),
     createBoolConfig("soft-shutdown", NULL, MODIFIABLE_CONFIG, g_pserver->config_soft_shutdown, 0, NULL, NULL),
+    createSizeTConfig("semi-ordered-set-bucket-size", NULL, MODIFIABLE_CONFIG, 0, 1024, g_semiOrderedSetTargetBucketSize, 0, INTEGER_CONFIG, NULL, NULL),
 
 #ifdef USE_OPENSSL
     createIntConfig("tls-port", NULL, MODIFIABLE_CONFIG, 0, 65535, g_pserver->tls_port, 0, INTEGER_CONFIG, NULL, updateTLSPort), /* TCP port. */

src/db.cpp

@@ -93,7 +93,7 @@ static void lookupKeyUpdateObj(robj *val, int flags)
     /* Update the access time for the ageing algorithm.
      * Don't do it if we have a saving child, as this will trigger
      * a copy on write madness. */
-    if (!hasActiveChildProcess() && !(flags & LOOKUP_NOTOUCH))
+    if (!hasActiveChildProcessOrBGSave() && !(flags & LOOKUP_NOTOUCH))
     {
         if (g_pserver->maxmemory_policy & MAXMEMORY_FLAG_LFU) {
             updateLFU(val);
@@ -1059,7 +1059,7 @@ void keysCommand(client *c) {
     sds pattern = szFromObj(c->argv[1]);
 
     const redisDbPersistentDataSnapshot *snapshot = nullptr;
-    if (!(c->flags & (CLIENT_MULTI | CLIENT_BLOCKED)))
+    if (!(c->flags & (CLIENT_MULTI | CLIENT_BLOCKED | CLIENT_DENY_BLOCKING)))
         snapshot = c->db->createSnapshot(c->mvccCheckpoint, true /* fOptional */);
     if (snapshot != nullptr)
     {
@@ -1737,6 +1737,9 @@ int dbSwapDatabases(int id1, int id2) {
         id2 < 0 || id2 >= cserver.dbnum) return C_ERR;
     if (id1 == id2) return C_OK;
     std::swap(g_pserver->db[id1], g_pserver->db[id2]);
+    
+    //swap db's id too, otherwise db does not match its id
+    std::swap(g_pserver->db[id1]->id, g_pserver->db[id2]->id);
 
     /* Note that we don't swap blocking_keys,
      * ready_keys and watched_keys, since we want clients to
@@ -1766,7 +1769,7 @@ int dbSwapDatabases(int id1, int id2) {
 
 /* SWAPDB db1 db2 */
 void swapdbCommand(client *c) {
-    int id1, id2;
+    int id1, id2, oriId;
 
     /* Not allowed in cluster mode: we have just DB 0 there. */
     if (g_pserver->cluster_enabled) {
@@ -1783,6 +1786,9 @@ void swapdbCommand(client *c) {
         "invalid second DB index") != C_OK)
         return;
 
+    //get client's original db's id
+    oriId=c->db->id;
+
     /* Swap... */
     if (dbSwapDatabases(id1,id2) == C_ERR) {
         addReplyError(c,"DB index is out of range");
@@ -1791,6 +1797,10 @@ void swapdbCommand(client *c) {
         RedisModuleSwapDbInfo si = {REDISMODULE_SWAPDBINFO_VERSION,(int32_t)id1,(int32_t)id2};
         moduleFireServerEvent(REDISMODULE_EVENT_SWAPDB,0,&si);
         g_pserver->dirty++;
+
+	//set client's db to original db
+        c->db=g_pserver->db[oriId];
+
         addReply(c,shared.ok);
     }
 }
@@ -1931,6 +1941,7 @@ void setExpire(client *c, redisDb *db, robj *key, expireEntry &&e)
  * is associated with this key (i.e. the key is non volatile) */
 expireEntry *redisDbPersistentDataSnapshot::getExpire(const char *key) {
     /* No expire? return ASAP */
+    std::unique_lock<fastlock> ul(g_expireLock);
     if (expireSize() == 0)
         return nullptr;
 
@@ -2607,6 +2618,17 @@ void clusterStorageLoadCallback(const char *rgchkey, size_t cch, void *)
     slotToKeyUpdateKeyCore(rgchkey, cch, true /*add*/);
 }
 
+void moduleLoadCallback(const char * rgchKey, size_t, void *data) {
+    redisObjectStack keyobj;
+    initStaticStringObject(keyobj, const_cast<char *>(rgchKey));
+    moduleNotifyKeyspaceEvent(NOTIFY_LOADED, "loaded", &keyobj, *(int *)data);
+}
+
+void moduleClusterLoadCallback(const char * rgchKey, size_t cchKey, void *data) {
+    clusterStorageLoadCallback(rgchKey, cchKey, data);
+    moduleLoadCallback(rgchKey, cchKey, data);
+}
+
 void redisDb::initialize(int id)
 {
     redisDbPersistentData::initialize();
@@ -2625,8 +2647,8 @@ void redisDb::storageProviderInitialize()
 {
     if (g_pserver->m_pstorageFactory != nullptr)
     {
-        IStorageFactory::key_load_iterator itr = (g_pserver->cluster_enabled) ? clusterStorageLoadCallback : nullptr;
-        this->setStorageProvider(StorageCache::create(g_pserver->m_pstorageFactory, id, itr, nullptr));
+        IStorageFactory::key_load_iterator itr = g_pserver->cluster_enabled ? moduleClusterLoadCallback : moduleLoadCallback;
+        this->setStorageProvider(StorageCache::create(g_pserver->m_pstorageFactory, id, itr, &id));
     }
 }
 
@@ -2716,8 +2738,11 @@ void redisDbPersistentData::clear(void(callback)(void*))
         m_cnewKeysPending = 0;
         m_fAllChanged++;
     }
+    {
+    std::unique_lock<fastlock> ul(g_expireLock);
     delete m_setexpire;
     m_setexpire = new (MALLOC_LOCAL) expireset();
+    }
     if (m_spstorage != nullptr)
         m_spstorage->clear(callback);
     dictEmpty(m_pdictTombstone,callback);
@@ -2850,7 +2875,8 @@ LNotFound:
             {
                 dictAdd(m_pdict, sdsNewKey, o);
                 o->SetFExpires(spexpire != nullptr);
-
+                
+                std::unique_lock<fastlock> ul(g_expireLock);
                 if (spexpire != nullptr)
                 {
                     auto itr = m_setexpire->find(sdsKey);
@@ -2873,6 +2899,7 @@ LNotFound:
     if (*pde != nullptr && dictGetVal(*pde) != nullptr)
     {
         robj *o = (robj*)dictGetVal(*pde);
+        std::unique_lock<fastlock> ul(g_expireLock);
         serverAssert(o->FExpires() == (m_setexpire->find(sdsKey) != m_setexpire->end()));
     }
 }
@@ -2994,8 +3021,16 @@ void redisDbPersistentData::processChangesAsync(std::atomic<int> &pendingJobs)
     });
 }
 
-void redisDbPersistentData::bulkStorageInsert(char **rgKeys, size_t *rgcbKeys, char **rgVals, size_t *rgcbVals, size_t celem)
+/* This function is to bulk insert directly to storage provider bypassing in memory, assumes rgKeys and rgVals are not sds strings */
+void redisDbPersistentData::bulkDirectStorageInsert(char **rgKeys, size_t *rgcbKeys, char **rgVals, size_t *rgcbVals, size_t celem)
 {
+    if (g_pserver->cluster_enabled) {
+        aeAcquireLock();
+        for (size_t i = 0; i < celem; i++) {
+            slotToKeyUpdateKeyCore(rgKeys[i], rgcbKeys[i], 1);
+        }
+        aeReleaseLock();
+    }
     m_spstorage->bulkInsert(rgKeys, rgcbKeys, rgVals, rgcbVals, celem);
 }
 
@@ -3237,6 +3272,7 @@ std::unique_ptr<expireEntry> deserializeExpire(sds key, const char *str, size_t
 
 sds serializeStoredObjectAndExpire(redisDbPersistentData *db, const char *key, robj_roptr o)
 {
+    std::unique_lock<fastlock> ul(g_expireLock);
     auto itrExpire = db->setexpire()->find(key);
     const expireEntry *pexpire = nullptr;
     if (itrExpire != db->setexpire()->end())
@@ -3257,7 +3293,7 @@ int dbnumFromDb(redisDb *db)
     serverPanic("invalid database pointer");
 }
 
-bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command, bool fExecOK)
+void redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command)
 {
     if (m_spstorage == nullptr) {
 #if defined(__x86_64__) || defined(__i386__)
@@ -3269,7 +3305,7 @@ bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
             const char *cmd = szFromObj(command.argv[0]);
             if (!strcasecmp(cmd, "set") || !strcasecmp(cmd, "get")) {
                 if (c->db->m_spdbSnapshotHOLDER != nullptr)
-                    return false; // this is dangerous enough without a snapshot around
+                    return; // this is dangerous enough without a snapshot around
                 auto h = dictSdsHash(szFromObj(command.argv[1]));
                 for (int iht = 0; iht < 2; ++iht) {
                     auto hT = h & c->db->m_pdict->ht[iht].sizemask;
@@ -3289,7 +3325,7 @@ bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
             }
         }
 #endif
-        return false;
+        return;
     }
 
     AeLocker lock;
@@ -3299,10 +3335,10 @@ bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
     getKeysResult result = GETKEYS_RESULT_INIT;
     auto cmd = lookupCommand(szFromObj(command.argv[0]));
     if (cmd == nullptr)
-        return false; // Bad command? It's not for us to judge, just bail
+        return; // Bad command? It's not for us to judge, just bail
     
     if (command.argc < std::abs(cmd->arity))
-        return false; // Invalid number of args
+        return; // Invalid number of args
     
     int numkeys = getKeysFromCommand(cmd, command.argv, command.argc, &result);
     for (int ikey = 0; ikey < numkeys; ++ikey)
@@ -3335,7 +3371,6 @@ bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
         }
     }
 
-    bool fNoInsert = false;
     if (!vecInserts.empty()) {
         lock.arm(c);
         for (auto &tuple : vecInserts)
@@ -3351,19 +3386,18 @@ bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
                     // While unlocked this was already ensured
                     decrRefCount(o);
                     sdsfree(sharedKey);
-                    fNoInsert = true;
                 }
                 else
                 {
                     if (spexpire != nullptr) {
                         if (spexpire->when() < mstime()) {
-                            fNoInsert = true;
                             break;
                         }
                     }
                     dictAdd(m_pdict, sharedKey, o);
                     o->SetFExpires(spexpire != nullptr);
 
+                    std::unique_lock<fastlock> ul(g_expireLock);
                     if (spexpire != nullptr)
                     {
                         auto itr = m_setexpire->find(sharedKey);
@@ -3384,12 +3418,5 @@ bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
         lock.disarm();
     }
 
-    if (fExecOK && !fNoInsert && cmd->proc == getCommand && !vecInserts.empty()) {
-        robj *o = std::get<1>(vecInserts[0]);
-        if (o != nullptr) {
-            addReplyBulk(c, o);
-            return true;
-        }
-    }
-    return false;
-}
+    return;
+}

src/debug.cpp

@@ -341,10 +341,10 @@ void mallctl_int(client *c, robj **argv, int argc) {
     }
     size_t sz = sizeof(old);
     while (sz > 0) {
-        if ((ret=je_mallctl(szFromObj(argv[0]), &old, &sz, argc > 1? &val: NULL, argc > 1?sz: 0))) {
+        if ((ret=mallctl(szFromObj(argv[0]), &old, &sz, argc > 1? &val: NULL, argc > 1?sz: 0))) {
             if (ret == EPERM && argc > 1) {
                 /* if this option is write only, try just writing to it. */
-                if (!(ret=je_mallctl(szFromObj(argv[0]), NULL, 0, &val, sz))) {
+                if (!(ret=mallctl(szFromObj(argv[0]), NULL, 0, &val, sz))) {
                     addReply(c, shared.ok);
                     return;
                 }
@@ -375,7 +375,7 @@ void mallctl_string(client *c, robj **argv, int argc) {
     char *old;
     size_t sz = sizeof(old);
     /* for strings, it seems we need to first get the old value, before overriding it. */
-    if ((rret=je_mallctl(szFromObj(argv[0]), &old, &sz, NULL, 0))) {
+    if ((rret=mallctl(szFromObj(argv[0]), &old, &sz, NULL, 0))) {
         /* return error unless this option is write only. */
         if (!(rret == EPERM && argc > 1)) {
             addReplyErrorFormat(c,"%s", strerror(rret));
@@ -387,7 +387,7 @@ void mallctl_string(client *c, robj **argv, int argc) {
         char **valref = &val;
         if ((!strcmp(val,"VOID")))
             valref = NULL, sz = 0;
-        wret = je_mallctl(szFromObj(argv[0]), NULL, 0, valref, sz);
+        wret = mallctl(szFromObj(argv[0]), NULL, 0, valref, sz);
     }
     if (!rret)
         addReplyBulkCString(c, old);

src/dict.cpp

@@ -1577,7 +1577,7 @@ char *stringFromLongLong(long long value) {
     int len;
     char *s;
 
-    len = snprintf(buf,32,"%lld",value);
+    len = snprintf(buf,sizeof(buf),"%lld",value);
     s = zmalloc(len+1);
     memcpy(s, buf, len);
     s[len] = '\0';

src/endianconv.c

@@ -112,15 +112,15 @@ int endianconvTest(int argc, char *argv[], int accurate) {
     UNUSED(argv);
     UNUSED(accurate);
 
-    snprintf(buf,32,"ciaoroma");
+    snprintf(buf,sizeof(buf),"ciaoroma");
     memrev16(buf);
     printf("%s\n", buf);
 
-    snprintf(buf,32,"ciaoroma");
+    snprintf(buf,sizeof(buf),"ciaoroma");
     memrev32(buf);
     printf("%s\n", buf);
 
-    snprintf(buf,32,"ciaoroma");
+    snprintf(buf,sizeof(buf),"ciaoroma");
     memrev64(buf);
     printf("%s\n", buf);
 

src/evict.cpp

@@ -380,7 +380,7 @@ size_t freeMemoryGetNotCountedMemory(void) {
 
     /* also don't count the replication backlog memory
      * that's where the replication clients get their memory from */
-    overhead += g_pserver->repl_backlog_size;
+    overhead += g_pserver->repl_backlog_size - g_pserver->repl_backlog_config_size;
 
     if (g_pserver->aof_state != AOF_OFF) {
         overhead += sdsalloc(g_pserver->aof_buf)+aofRewriteBufferSize();
@@ -771,6 +771,7 @@ int performEvictions(bool fPreSnapshot) {
                 if (db->removeCachedValue(bestkey, &deT)) {
                     mem_freed += splazy->addEntry(db->dictUnsafeKeyOnly(), deT);
                     ckeysFailed = 0;
+                    g_pserver->stat_evictedkeys++;
                 }
                 else {
                     delta = 0;

src/expire.cpp

@@ -416,6 +416,33 @@ void activeExpireCycleCore(int type) {
             return true;
         }, &check);
 
+        if (!timelimit_exit && db->getStorageCache() != nullptr)
+            db->getStorageCache()->stateful_enumerate([&] (const char *key, size_t keylen, const void *val, size_t vallen) {
+                sds sdsKey = sdsnewlen(key, keylen);
+                size_t offset = 0;
+                auto e = deserializeExpire(sdsKey, (const char*)val, vallen, &offset);
+                if (e->when() < now)
+                {
+                    expired += activeExpireCycleExpire(db, *e, now, tried);
+                }
+                sdsfree(sdsKey);
+
+                if ((tried % ACTIVE_EXPIRE_CYCLE_LOOKUPS_PER_LOOP) == 0)
+                {
+                    /* We can't block forever here even if there are many keys to
+                    * expire. So after a given amount of milliseconds return to the
+                    * caller waiting for the other active expire cycle. */
+                    elapsed = ustime()-start;
+                    if (elapsed > timelimit) {
+                        timelimit_exit = 1;
+                        g_pserver->stat_expired_time_cap_reached_count++;
+                        return false;
+                    }
+                    check = ACTIVE_EXPIRE_CYCLE_FAST_DURATION;
+                }
+                return true;
+            });
+
         total_expired += expired;
     }
 
@@ -481,6 +508,7 @@ void expireSlaveKeys(void) {
     if (slaveKeysWithExpire == NULL ||
         dictSize(slaveKeysWithExpire) == 0) return;
 
+    std::unique_lock<fastlock> ul(g_expireLock);
     int cycles = 0, noexpire = 0;
     mstime_t start = mstime();
     while(1) {

src/keydb-diagnostic-tool.cpp

@@ -904,7 +904,7 @@ int main(int argc, const char **argv) {
 
     while (self_threads < config.max_threads) {
         for (int i = 0; i < config.numclients; i++) {
-            snprintf(command, 63, "SET %d %s\r\n", self_threads * config.numclients + i, set_value);
+            snprintf(command, sizeof(command), "SET %d %s\r\n", self_threads * config.numclients + i, set_value);
             createClient(command, strlen(command), NULL,self_threads);
         }
 

src/modules/hellotimer.c

@@ -51,8 +51,9 @@ int TimerCommand_RedisCommand(RedisModuleCtx *ctx, RedisModuleString **argv, int
 
     for (int j = 0; j < 10; j++) {
         int delay = rand() % 5000;
-        char *buf = RedisModule_Alloc(256);
-        snprintf(buf,256,"After %d", delay);
+        int bufsize = 256;
+        char *buf = RedisModule_Alloc(bufsize);
+        snprintf(buf,bufsize,"After %d", delay);
         RedisModuleTimerID tid = RedisModule_CreateTimer(ctx,delay,timerHandler,buf);
         REDISMODULE_NOT_USED(tid);
     }

src/networking.cpp

@@ -145,7 +145,7 @@ client *createClient(connection *conn, int iel) {
     client_id = g_pserver->next_client_id.fetch_add(1);
     c->iel = iel;
     c->id = client_id;
-    snprintf(c->lock.szName, 56, "client %" PRIu64, client_id);
+    snprintf(c->lock.szName, sizeof(c->lock.szName), "client %" PRIu64, client_id);
     c->resp = 2;
     c->conn = conn;
     c->name = NULL;
@@ -777,11 +777,11 @@ void setDeferredAggregateLen(client *c, void *node, long length, char prefix) {
          * we return NULL in addReplyDeferredLen() */
         if (node == NULL) return;
         char lenstr[128];
-        size_t lenstr_len = snprintf(lenstr, 128, "%c%ld\r\n", prefix, length);
+        size_t lenstr_len = snprintf(lenstr, sizeof(lenstr), "%c%ld\r\n", prefix, length);
         setDeferredReply(c, node, lenstr, lenstr_len);
     } else {
         char lenstr[128];
-        int lenstr_len = snprintf(lenstr, 128, "%c%ld\r\n", prefix, length);
+        int lenstr_len = snprintf(lenstr, sizeof(lenstr), "%c%ld\r\n", prefix, length);
 
         size_t idxSplice = (size_t)node;
         serverAssert(idxSplice <= c->replyAsync->used);
@@ -2595,9 +2595,7 @@ void parseClientCommandBuffer(client *c) {
             (g_pserver->m_pstorageFactory || aeLockContested(cserver.cthreads/2) || cserver.cthreads == 1) && !GlobalLocksAcquired()) {
             auto &query = c->vecqueuedcmd.back();
             if (query.argc > 0 && query.argc == query.argcMax) {
-                if (c->db->prefetchKeysAsync(c, query, c->vecqueuedcmd.size() == 1)) {
-                    c->vecqueuedcmd.erase(c->vecqueuedcmd.begin());
-                }
+                c->db->prefetchKeysAsync(c, query);
             }
         }
         c->reqtype = 0;

src/new.cpp

@@ -41,10 +41,3 @@ void operator delete(void *p, std::size_t) noexcept
 }
 
 #endif
-
-#if defined(USE_JEMALLOC)
-extern "C" size_t malloc_usable_size(void *ptr)
-{
-    return zmalloc_usable_size(ptr);
-}
-#endif

src/object.cpp

@@ -1541,7 +1541,7 @@ void memoryCommand(client *c) {
     } else if (!strcasecmp(szFromObj(c->argv[1]),"malloc-stats") && c->argc == 2) {
 #if defined(USE_JEMALLOC)
         sds info = sdsempty();
-        je_malloc_stats_print(inputCatSds, &info, NULL);
+        malloc_stats_print(inputCatSds, &info, NULL);
         addReplyVerbatim(c,info,sdslen(info),"txt");
         sdsfree(info);
 #else
@@ -1668,6 +1668,7 @@ robj *deserializeStoredObjectCore(const void *data, size_t cb)
 robj *deserializeStoredObject(const redisDbPersistentData *db, const char *key, const void *data, size_t cb)
 {
     robj *o = deserializeStoredObjectCore(data, cb);
+    std::unique_lock<fastlock> ul(g_expireLock);
     o->SetFExpires(db->setexpire()->exists(key));
     return o;
 }

src/rdb.cpp

@@ -1163,7 +1163,7 @@ int rdbSaveKeyValuePair(rio *rdb, robj_roptr key, robj_roptr val, const expireEn
 
     char szT[32];
     if (g_pserver->fActiveReplica) {
-        snprintf(szT, 32, "%" PRIu64, mvccFromObj(val));
+        snprintf(szT, sizeof(szT), "%" PRIu64, mvccFromObj(val));
         if (rdbSaveAuxFieldStrStr(rdb,"mvcc-tstamp", szT) == -1) return -1;
     }
 
@@ -1190,7 +1190,7 @@ int rdbSaveKeyValuePair(rio *rdb, robj_roptr key, robj_roptr val, const expireEn
         {
             if (itr.subkey() == nullptr)
                 continue;   // already saved
-            snprintf(szT, 32, "%lld", itr.when());
+            snprintf(szT, sizeof(szT), "%lld", itr.when());
             rdbSaveAuxFieldStrStr(rdb,"keydb-subexpire-key",itr.subkey());
             rdbSaveAuxFieldStrStr(rdb,"keydb-subexpire-when",szT);
         }
@@ -1334,7 +1334,7 @@ int rdbSaveRio(rio *rdb, const redisDbPersistentDataSnapshot **rgpdb, int *error
     if (rdbSaveModulesAux(rdb, REDISMODULE_AUX_BEFORE_RDB) == -1) goto werr;
 
     for (j = 0; j < cserver.dbnum; j++) {
-        const redisDbPersistentDataSnapshot *db = rgpdb[j];
+        const redisDbPersistentDataSnapshot *db = rgpdb != nullptr ? rgpdb[j] : g_pserver->db[j];
         if (db->size() == 0) continue;
 
         /* Write the SELECT DB opcode */
@@ -1490,7 +1490,7 @@ int rdbSaveFile(char *filename, const redisDbPersistentDataSnapshot **rgpdb, rdb
     rio rdb;
     int error = 0;
 
-    snprintf(tmpfile,256,"temp-%d-%d.rdb", getpid(), g_pserver->rdbThreadVars.tmpfileNum);
+    getTempFileName(tmpfile, g_pserver->rdbThreadVars.tmpfileNum);
     fp = fopen(tmpfile,"w");
     if (!fp) {
         char *cwdp = getcwd(cwd,MAXPATHLEN);
@@ -1660,12 +1660,14 @@ int launchRdbSaveThread(pthread_t &child, rdbSaveInfo *rsi)
         pthread_attr_t tattr;
         pthread_attr_init(&tattr);
         pthread_attr_setstacksize(&tattr, 1 << 23); // 8 MB
+        openChildInfoPipe();
         if (pthread_create(&child, &tattr, rdbSaveThread, args)) {
             pthread_attr_destroy(&tattr);
             for (int idb = 0; idb < cserver.dbnum; ++idb)
                 g_pserver->db[idb]->endSnapshot(args->rgpdb[idb]);
             args->~rdbSaveThreadArgs();
             zfree(args);
+            closeChildInfoPipe();
             return C_ERR;
         }
         pthread_attr_destroy(&tattr);
@@ -1679,24 +1681,23 @@ int rdbSaveBackground(rdbSaveInfo *rsi) {
     pthread_t child;
     long long start;
 
-    if (hasActiveChildProcess()) return C_ERR;
+    if (hasActiveChildProcessOrBGSave()) return C_ERR;
 
     g_pserver->dirty_before_bgsave = g_pserver->dirty;
     g_pserver->lastbgsave_try = time(NULL);
-    openChildInfoPipe();
 
     start = ustime();
+    latencyStartMonitor(g_pserver->rdb_save_latency);
 
-    
-    g_pserver->stat_fork_time = ustime()-start;
-    g_pserver->stat_fork_rate = (double) zmalloc_used_memory() * 1000000 / g_pserver->stat_fork_time / (1024*1024*1024); /* GB per second. */
     if (launchRdbSaveThread(child, rsi) != C_OK) {
-        closeChildInfoPipe();
         g_pserver->lastbgsave_status = C_ERR;
         serverLog(LL_WARNING,"Can't save in background: fork: %s",
             strerror(errno));
         return C_ERR;
     }
+
+    g_pserver->stat_fork_time = ustime()-start;
+    g_pserver->stat_fork_rate = (double) zmalloc_used_memory() * 1000000 / g_pserver->stat_fork_time / (1024*1024*1024); /* GB per second. */
     latencyAddSampleIfNeeded("fork",g_pserver->stat_fork_time/1000);
     serverLog(LL_NOTICE,"Background saving started");
     g_pserver->rdb_save_time_start = time(NULL);
@@ -1714,7 +1715,7 @@ void getTempFileName(char tmpfile[], int tmpfileNum) {
     char tmpfileNumString[214];
 
     /* Generate temp rdb file name using aync-signal safe functions. */
-    int pid_len = ll2string(pid, sizeof(pid), getpid());
+    int pid_len = ll2string(pid, sizeof(pid), g_pserver->in_fork_child ? getppid() : getpid());
     int tmpfileNum_len = ll2string(tmpfileNumString, sizeof(tmpfileNumString), tmpfileNum);
     strcpy(tmpfile, "temp-");
     strncpy(tmpfile+5, pid, pid_len);
@@ -3601,6 +3602,8 @@ static void backgroundSaveDoneHandlerDisk(int exitcode, bool fCancelled) {
         g_pserver->dirty = g_pserver->dirty - g_pserver->dirty_before_bgsave;
         g_pserver->lastsave = time(NULL);
         g_pserver->lastbgsave_status = C_OK;
+        latencyEndMonitor(g_pserver->rdb_save_latency);
+        latencyAddSampleIfNeeded("rdb-save",g_pserver->rdb_save_latency);
     } else if (!fCancelled && exitcode != 0) {
         serverLog(LL_WARNING, "Background saving error");
         g_pserver->lastbgsave_status = C_ERR;
@@ -3691,7 +3694,7 @@ void killRDBChild(bool fSynchronous) {
     serverAssert(GlobalLocksAcquired());
 
     if (cserver.fForkBgSave) {
-        kill(g_pserver->rdb_child_pid,SIGUSR1);
+        kill(g_pserver->child_pid,SIGUSR1);
     } else { 
         g_pserver->rdbThreadVars.fRdbThreadCancel = true;
         if (g_pserver->rdb_child_type == RDB_CHILD_TYPE_SOCKET) {
@@ -3782,7 +3785,7 @@ int rdbSaveToSlavesSockets(rdbSaveInfo *rsi) {
     int pipefds[2];
     rdbSaveSocketThreadArgs *args = nullptr;
 
-    if (hasActiveChildProcess()) return C_ERR;
+    if (hasActiveChildProcessOrBGSave()) return C_ERR;
 
     /* Even if the previous fork child exited, don't start a new one until we
      * drained the pipe. */
@@ -3829,57 +3832,130 @@ int rdbSaveToSlavesSockets(rdbSaveInfo *rsi) {
     }
 
     /* Create the child process. */
-    openChildInfoPipe();
+    if (cserver.fForkBgSave) {
+        pid_t childpid;
+        if ((childpid = redisFork(CHILD_TYPE_RDB)) == 0) {
+            /* Child */
+            int retval, dummy;
+            rio rdb;
 
-    for (int idb = 0; idb < cserver.dbnum; ++idb)
-        args->rgpdb[idb] = g_pserver->db[idb]->createSnapshot(getMvccTstamp(), false /*fOptional*/);
+            rioInitWithFd(&rdb,args->rdb_pipe_write);
 
-    g_pserver->rdbThreadVars.tmpfileNum++;
-    g_pserver->rdbThreadVars.fRdbThreadCancel = false;
-    pthread_attr_t tattr;
-    pthread_attr_init(&tattr);
-    pthread_attr_setstacksize(&tattr, 1 << 23); // 8 MB
-    if (pthread_create(&child, &tattr, rdbSaveToSlavesSocketsThread, args)) {
-        pthread_attr_destroy(&tattr);
-        serverLog(LL_WARNING,"Can't save in background: fork: %s",
-            strerror(errno));
+            redisSetProcTitle("keydb-rdb-to-slaves");
+            redisSetCpuAffinity(g_pserver->bgsave_cpulist);
 
-        /* Undo the state change. The caller will perform cleanup on
-            * all the slaves in BGSAVE_START state, but an early call to
-            * replicationSetupSlaveForFullResync() turned it into BGSAVE_END */
-        listRewind(g_pserver->slaves,&li);
-        while((ln = listNext(&li))) {
-            client *replica = (client*)ln->value;
-            if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_END) {
-                replica->replstate = SLAVE_STATE_WAIT_BGSAVE_START;
+            retval = rdbSaveRioWithEOFMark(&rdb,nullptr,nullptr,rsi);
+            if (retval == C_OK && rioFlush(&rdb) == 0)
+                retval = C_ERR;
+
+            if (retval == C_OK) {
+                sendChildCowInfo(CHILD_INFO_TYPE_RDB_COW_SIZE, "RDB");
             }
-        }
-        close(args->rdb_pipe_write);
-        close(g_pserver->rdb_pipe_read);
-        zfree(g_pserver->rdb_pipe_conns);
-        close(args->safe_to_exit_pipe);
-        g_pserver->rdb_pipe_conns = NULL;
-        g_pserver->rdb_pipe_numconns = 0;
-        g_pserver->rdb_pipe_numconns_writing = 0;
-        args->rsi.~rdbSaveInfo();
-        zfree(args);
-        closeChildInfoPipe();
-        return C_ERR;
-    }
-    pthread_attr_destroy(&tattr);
-    g_pserver->child_type = CHILD_TYPE_RDB;
 
-    serverLog(LL_NOTICE,"Background RDB transfer started");
-    g_pserver->rdb_save_time_start = time(NULL);
-    serverAssert(!g_pserver->rdbThreadVars.fRdbThreadActive);
-    g_pserver->rdbThreadVars.rdb_child_thread = child;
-    g_pserver->rdbThreadVars.fRdbThreadActive = true;
-    g_pserver->rdb_child_type = RDB_CHILD_TYPE_SOCKET;
-    aePostFunction(g_pserver->rgthreadvar[IDX_EVENT_LOOP_MAIN].el, []{
-        if (aeCreateFileEvent(serverTL->el, g_pserver->rdb_pipe_read, AE_READABLE, rdbPipeReadHandler,NULL) == AE_ERR) {
-            serverPanic("Unrecoverable error creating server.rdb_pipe_read file event.");
+            rioFreeFd(&rdb);
+            /* wake up the reader, tell it we're done. */
+            close(args->rdb_pipe_write);
+            close(g_pserver->rdb_child_exit_pipe); /* close write end so that we can detect the close on the parent. */
+            /* hold exit until the parent tells us it's safe. we're not expecting
+            * to read anything, just get the error when the pipe is closed. */
+            dummy = read(args->safe_to_exit_pipe, pipefds, 1);
+            UNUSED(dummy);
+            exitFromChild((retval == C_OK) ? 0 : 1);
+        } else {
+            /* Parent */
+            close(args->safe_to_exit_pipe);
+            if (childpid == -1) {
+                serverLog(LL_WARNING,"Can't save in background: fork: %s",
+                    strerror(errno));
+
+                /* Undo the state change. The caller will perform cleanup on
+                * all the slaves in BGSAVE_START state, but an early call to
+                * replicationSetupSlaveForFullResync() turned it into BGSAVE_END */
+                listRewind(g_pserver->slaves,&li);
+                while((ln = listNext(&li))) {
+                    client *replica = (client*)ln->value;
+                    if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_END) {
+                        replica->replstate = SLAVE_STATE_WAIT_BGSAVE_START;
+                    }
+                }
+                close(args->rdb_pipe_write);
+                close(g_pserver->rdb_pipe_read);
+                zfree(g_pserver->rdb_pipe_conns);
+                g_pserver->rdb_pipe_conns = NULL;
+                g_pserver->rdb_pipe_numconns = 0;
+                g_pserver->rdb_pipe_numconns_writing = 0;
+                args->rsi.~rdbSaveInfo();
+                zfree(args);
+            } else {
+                serverLog(LL_NOTICE,"Background RDB transfer started by pid %ld",
+                    (long)childpid);
+                g_pserver->rdb_save_time_start = time(NULL);
+                g_pserver->rdb_child_type = RDB_CHILD_TYPE_SOCKET;
+                g_pserver->rdbThreadVars.fRdbThreadActive = true;
+                updateDictResizePolicy();
+                close(args->rdb_pipe_write); /* close write in parent so that it can detect the close on the child. */
+                aePostFunction(g_pserver->rgthreadvar[IDX_EVENT_LOOP_MAIN].el, []{
+                    if (aeCreateFileEvent(serverTL->el, g_pserver->rdb_pipe_read, AE_READABLE, rdbPipeReadHandler, nullptr) == AE_ERR) {
+                        serverPanic("Unrecoverable error creating g_pserver->rdb_pipe_read file event.");
+                    }
+                });
+            }
+            return (childpid == -1) ? C_ERR : C_OK;
         }
-    });
+    }
+    else {
+        openChildInfoPipe();
+
+        for (int idb = 0; idb < cserver.dbnum; ++idb)
+            args->rgpdb[idb] = g_pserver->db[idb]->createSnapshot(getMvccTstamp(), false /*fOptional*/);
+
+        g_pserver->rdbThreadVars.tmpfileNum++;
+        g_pserver->rdbThreadVars.fRdbThreadCancel = false;
+        pthread_attr_t tattr;
+        pthread_attr_init(&tattr);
+        pthread_attr_setstacksize(&tattr, 1 << 23); // 8 MB
+        if (pthread_create(&child, &tattr, rdbSaveToSlavesSocketsThread, args)) {
+            pthread_attr_destroy(&tattr);
+            serverLog(LL_WARNING,"Can't save in background: fork: %s",
+                strerror(errno));
+
+            /* Undo the state change. The caller will perform cleanup on
+                * all the slaves in BGSAVE_START state, but an early call to
+                * replicationSetupSlaveForFullResync() turned it into BGSAVE_END */
+            listRewind(g_pserver->slaves,&li);
+            while((ln = listNext(&li))) {
+                client *replica = (client*)ln->value;
+                if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_END) {
+                    replica->replstate = SLAVE_STATE_WAIT_BGSAVE_START;
+                }
+            }
+            close(args->rdb_pipe_write);
+            close(g_pserver->rdb_pipe_read);
+            zfree(g_pserver->rdb_pipe_conns);
+            close(args->safe_to_exit_pipe);
+            g_pserver->rdb_pipe_conns = NULL;
+            g_pserver->rdb_pipe_numconns = 0;
+            g_pserver->rdb_pipe_numconns_writing = 0;
+            args->rsi.~rdbSaveInfo();
+            zfree(args);
+            closeChildInfoPipe();
+            return C_ERR;
+        }
+        pthread_attr_destroy(&tattr);
+        g_pserver->child_type = CHILD_TYPE_RDB;
+
+        serverLog(LL_NOTICE,"Background RDB transfer started");
+        g_pserver->rdb_save_time_start = time(NULL);
+        serverAssert(!g_pserver->rdbThreadVars.fRdbThreadActive);
+        g_pserver->rdbThreadVars.rdb_child_thread = child;
+        g_pserver->rdbThreadVars.fRdbThreadActive = true;
+        g_pserver->rdb_child_type = RDB_CHILD_TYPE_SOCKET;
+        aePostFunction(g_pserver->rgthreadvar[IDX_EVENT_LOOP_MAIN].el, []{
+            if (aeCreateFileEvent(serverTL->el, g_pserver->rdb_pipe_read, AE_READABLE, rdbPipeReadHandler, nullptr) == AE_ERR) {
+                serverPanic("Unrecoverable error creating server.rdb_pipe_read file event.");
+            }
+        });
+    }
 
     return C_OK; /* Unreached. */
 }

src/rdb.h

@@ -151,6 +151,7 @@ int rdbLoad(rdbSaveInfo *rsi, int rdbflags);
 int rdbLoadFile(const char *filename, rdbSaveInfo *rsi, int rdbflags);
 int rdbSaveBackground(rdbSaveInfo *rsi);
 int rdbSaveToSlavesSockets(rdbSaveInfo *rsi);
+void getTempFileName(char tmpfile[], int tmpfileNum);
 void rdbRemoveTempFile(pid_t childpid, int from_signal);
 int rdbSave(const redisDbPersistentDataSnapshot **rgpdb, rdbSaveInfo *rsi);
 int rdbSaveFile(char *filename, const redisDbPersistentDataSnapshot **rgpdb, rdbSaveInfo *rsi);

src/readwritelock.h

@@ -8,6 +8,7 @@ class readWriteLock {
     int m_readCount = 0;
     int m_writeCount = 0;
     bool m_writeWaiting = false;
+    bool m_notify = true;
 public:
     readWriteLock(const char *name) : m_readLock(name), m_writeLock(name) {}
 
@@ -65,7 +66,8 @@ public:
     void releaseRead() {
         std::unique_lock<fastlock> rm(m_readLock);
         m_readCount--;
-        m_cv.notify_all();
+        if (m_notify)
+            m_cv.notify_all();
     }
 
     void releaseWrite(bool exclusive = true) {
@@ -74,7 +76,8 @@ public:
         if (exclusive)
             m_writeLock.unlock();
         m_writeCount--;
-        m_cv.notify_all();
+        if (m_notify)
+            m_cv.notify_all();
     }
 
     void downgradeWrite(bool exclusive = true) {
@@ -82,6 +85,10 @@ public:
         acquireRead();
     }
 
+    void setNotify(bool notify) {
+        m_notify = notify;
+    }
+
     bool hasReader() {
         return m_readCount > 0;
     }

src/redis-cli.c

@@ -3293,8 +3293,8 @@ static redisReply *clusterManagerMigrateKeysInReply(clusterManagerNode *source,
     argv_len = zcalloc(argc * sizeof(size_t), MALLOC_LOCAL);
     char portstr[255];
     char timeoutstr[255];
-    snprintf(portstr, 10, "%d", target->port);
-    snprintf(timeoutstr, 10, "%d", timeout);
+    snprintf(portstr, sizeof(portstr), "%d", target->port);
+    snprintf(timeoutstr, sizeof(timeoutstr), "%d", timeout);
     argv[0] = "MIGRATE";
     argv_len[0] = 7;
     argv[1] = target->ip;
@@ -6825,7 +6825,7 @@ static long getLongInfoField(char *info, char *field) {
 
 /* Convert number of bytes into a human readable string of the form:
  * 100B, 2G, 100M, 4K, and so forth. */
-void bytesToHuman(char *s, long long n) {
+void bytesToHuman(char *s, long long n, size_t bufsize) {
     double d;
 
     if (n < 0) {
@@ -6835,17 +6835,17 @@ void bytesToHuman(char *s, long long n) {
     }
     if (n < 1024) {
         /* Bytes */
-        snprintf(s,256,"%lldB",n);
+        snprintf(s,bufsize,"%lldB",n);
         return;
     } else if (n < (1024*1024)) {
         d = (double)n/(1024);
-        snprintf(s,256,"%.2fK",d);
+        snprintf(s,bufsize,"%.2fK",d);
     } else if (n < (1024LL*1024*1024)) {
         d = (double)n/(1024*1024);
-        snprintf(s,256,"%.2fM",d);
+        snprintf(s,bufsize,"%.2fM",d);
     } else if (n < (1024LL*1024*1024*1024)) {
         d = (double)n/(1024LL*1024*1024);
-        snprintf(s,256,"%.2fG",d);
+        snprintf(s,bufsize,"%.2fG",d);
     }
 }
 
@@ -6875,38 +6875,38 @@ static void statMode(void) {
         for (j = 0; j < 20; j++) {
             long k;
 
-            snprintf(buf,64,"db%d:keys",j);
+            snprintf(buf,sizeof(buf),"db%d:keys",j);
             k = getLongInfoField(reply->str,buf);
             if (k == LONG_MIN) continue;
             aux += k;
         }
-        snprintf(buf,64,"%ld",aux);
+        snprintf(buf,sizeof(buf),"%ld",aux);
         printf("%-11s",buf);
 
         /* Used memory */
         aux = getLongInfoField(reply->str,"used_memory");
-        bytesToHuman(buf,aux);
+        bytesToHuman(buf,aux,sizeof(buf));
         printf("%-8s",buf);
 
         /* Clients */
         aux = getLongInfoField(reply->str,"connected_clients");
-        snprintf(buf,64,"%ld",aux);
+        snprintf(buf,sizeof(buf),"%ld",aux);
         printf(" %-8s",buf);
 
         /* Blocked (BLPOPPING) Clients */
         aux = getLongInfoField(reply->str,"blocked_clients");
-        snprintf(buf,64,"%ld",aux);
+        snprintf(buf,sizeof(buf),"%ld",aux);
         printf("%-8s",buf);
 
         /* Requests */
         aux = getLongInfoField(reply->str,"total_commands_processed");
-        snprintf(buf,64,"%ld (+%ld)",aux,requests == 0 ? 0 : aux-requests);
+        snprintf(buf,sizeof(buf),"%ld (+%ld)",aux,requests == 0 ? 0 : aux-requests);
         printf("%-19s",buf);
         requests = aux;
 
         /* Connections */
         aux = getLongInfoField(reply->str,"total_connections_received");
-        snprintf(buf,64,"%ld",aux);
+        snprintf(buf,sizeof(buf),"%ld",aux);
         printf(" %-12s",buf);
 
         /* Children */

src/redismodule.h

@@ -326,6 +326,7 @@ static const RedisModuleEvent
 #define REDISMODULE_SUBEVENT_LOADING_ENDED 3
 #define REDISMODULE_SUBEVENT_LOADING_FAILED 4
 #define _REDISMODULE_SUBEVENT_LOADING_NEXT 5
+#define REDISMODULE_SUBEVENT_LOADING_FLASH_START 6
 
 #define REDISMODULE_SUBEVENT_CLIENT_CHANGE_CONNECTED 0
 #define REDISMODULE_SUBEVENT_CLIENT_CHANGE_DISCONNECTED 1

src/replication.cpp

@@ -1538,7 +1538,7 @@ void syncCommand(client *c) {
         } else {
             /* We don't have a BGSAVE in progress, let's start one. Diskless
              * or disk-based mode is determined by replica's capacity. */
-            if (!hasActiveChildProcess()) {
+            if (!hasActiveChildProcessOrBGSave()) {
                 startBgsaveForReplication(c->slave_capa);
             } else {
                 serverLog(LL_NOTICE,
@@ -3736,7 +3736,7 @@ retry_connect:
         while(maxtries--) {
             auto dt = std::chrono::system_clock::now().time_since_epoch();
             auto dtMillisecond = std::chrono::duration_cast<std::chrono::milliseconds>(dt);
-            snprintf(tmpfile,256,
+            snprintf(tmpfile,sizeof(tmpfile),
                 "temp-%d.%ld.rdb",(int)dtMillisecond.count(),(long int)getpid());
             dfd = open(tmpfile,O_CREAT|O_WRONLY|O_EXCL,0644);
             if (dfd != -1) break;
@@ -3797,9 +3797,11 @@ int connectWithMaster(redisMaster *mi) {
     connSetPrivateData(mi->repl_transfer_s, mi);
     if (connConnect(mi->repl_transfer_s, mi->masterhost, mi->masterport,
                 NET_FIRST_BIND_ADDR, syncWithMaster) == C_ERR) {
-        int sev = g_pserver->enable_multimaster ? LL_NOTICE : LL_WARNING;   // with multimaster its not unheard of to intentiallionall have downed masters
-        serverLog(sev,"Unable to connect to MASTER: %s",
-                connGetLastError(mi->repl_transfer_s));
+        const char *err = "Unknown Error";
+        if (mi->repl_transfer_s->last_errno != 0)
+            err = connGetLastError(mi->repl_transfer_s);
+        int sev = g_pserver->enable_multimaster ? LL_NOTICE : LL_WARNING;   // with multimaster its not unheard of to intentionally have downed masters
+        serverLog(sev, "Unable to connect to MASTER: %s", err);
         connClose(mi->repl_transfer_s);
         mi->repl_transfer_s = NULL;
         return C_ERR;
@@ -4984,7 +4986,7 @@ void replicationStartPendingFork(void) {
      * In case of diskless replication, we make sure to wait the specified
      * number of seconds (according to configuration) so that other slaves
      * have the time to arrive before we start streaming. */
-    if (!hasActiveChildProcess()) {
+    if (!hasActiveChildProcessOrBGSave()) {
         time_t idle, max_idle = 0;
         int slaves_waiting = 0;
         int mincapa = -1;

src/scripting.cpp

@@ -1264,6 +1264,9 @@ void scriptingInit(int setup) {
 
     /* Finally set the table as 'redis' global var. */
     lua_setglobal(lua,"redis");
+    /* Set table as 'keydb' global var as well */
+    lua_getglobal(lua,"redis");
+    lua_setglobal(lua,"keydb");
 
     /* Replace math.random and math.randomseed with our implementations. */
     lua_getglobal(lua,"math");

src/semiorderedset.h

@@ -27,6 +27,7 @@ namespace keydbutils
     template<>
     size_t hash(const sdsview &);
 }
+extern size_t g_semiOrderedSetTargetBucketSize;
 
 template<typename T, typename T_KEY = T, bool MEMMOVE_SAFE = false>
 class semiorderedset
@@ -41,11 +42,14 @@ class semiorderedset
     size_t idxRehash = (1ULL << bits_min); 
     int cfPauseRehash = 0;
 
-    constexpr size_t targetElementsPerBucket()
+    inline size_t targetElementsPerBucket()
     {
         // Aim for roughly 4 cache lines per bucket (determined by imperical testing)
         //  lower values are faster but use more memory
-        return std::max((64/sizeof(T))*8, (size_t)2);
+        if (g_semiOrderedSetTargetBucketSize == 0)
+            return std::max((64/sizeof(T))*8, (size_t)2);
+        else
+            return g_semiOrderedSetTargetBucketSize;
     }
 
 public:

src/server.cpp

@@ -1706,7 +1706,7 @@ int redisDbPersistentData::incrementallyRehash() {
  * for dict.c to resize the hash tables accordingly to the fact we have an
  * active fork child running. */
 void updateDictResizePolicy(void) {
-    if (!hasActiveChildProcess() || (g_pserver->FRdbSaveInProgress() && !cserver.fForkBgSave))
+    if (!hasActiveChildProcess())
         dictEnableResize();
     else
         dictDisableResize();
@@ -1725,7 +1725,11 @@ const char *strChildType(int type) {
 /* Return true if there are active children processes doing RDB saving,
  * AOF rewriting, or some side process spawned by a loaded module. */
 int hasActiveChildProcess() {
-    return g_pserver->FRdbSaveInProgress() || g_pserver->child_pid != -1;
+    return g_pserver->child_pid != -1;
+}
+
+int hasActiveChildProcessOrBGSave() {
+    return g_pserver->FRdbSaveInProgress() || hasActiveChildProcess();
 }
 
 void resetChildState() {
@@ -2074,7 +2078,7 @@ void databasesCron(bool fMainThread) {
     /* Perform hash tables rehashing if needed, but only if there are no
      * other processes saving the DB on disk. Otherwise rehashing is bad
      * as will cause a lot of copy-on-write of memory pages. */
-    if (!hasActiveChildProcess() || g_pserver->FRdbSaveInProgress()) {
+    if (!hasActiveChildProcess()) {
         /* We use global counters so if we stop the computation at a given
          * DB we'll be able to start from the successive in the next
          * cron loop iteration. */
@@ -2215,6 +2219,7 @@ void checkChildrenDone(void) {
             g_pserver->rdbThreadVars.fRdbThreadCancel = false;
             g_pserver->rdbThreadVars.fDone = false;
             if (exitcode == 0) receiveChildInfo();
+            closeChildInfoPipe();
         }
     }
     else if ((pid = waitpid(-1, &statloc, WNOHANG)) != 0) {
@@ -2468,14 +2473,14 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
 
     /* Start a scheduled AOF rewrite if this was requested by the user while
      * a BGSAVE was in progress. */
-    if (!hasActiveChildProcess() &&
+    if (!hasActiveChildProcessOrBGSave() &&
         g_pserver->aof_rewrite_scheduled)
     {
         rewriteAppendOnlyFileBackground();
     }
 
     /* Check if a background saving or AOF rewrite in progress terminated. */
-    if (hasActiveChildProcess() || ldbPendingChildren())
+    if (hasActiveChildProcessOrBGSave() || ldbPendingChildren())
     {
         run_with_period(1000) receiveChildInfo();
         checkChildrenDone();
@@ -2517,7 +2522,7 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
 
         /* Trigger an AOF rewrite if needed. */
         if (g_pserver->aof_state == AOF_ON &&
-            !hasActiveChildProcess() &&
+            !hasActiveChildProcessOrBGSave() &&
             g_pserver->aof_rewrite_perc &&
             g_pserver->aof_current_size > g_pserver->aof_rewrite_min_size)
         {
@@ -2601,7 +2606,7 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
      * Note: this code must be after the replicationCron() call above so
      * make sure when refactoring this file to keep this order. This is useful
      * because we want to give priority to RDB savings for replication. */
-    if (!hasActiveChildProcess() &&
+    if (!hasActiveChildProcessOrBGSave() &&
         g_pserver->rdb_bgsave_scheduled &&
         (g_pserver->unixtime-g_pserver->lastbgsave_try > CONFIG_BGSAVE_RETRY_DELAY ||
          g_pserver->lastbgsave_status == C_OK))
@@ -2815,9 +2820,6 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
 
     runAndPropogateToReplicas(processClients);
 
-    /* Handle precise timeouts of blocked clients. */
-    handleBlockedClientsTimeout();
-
     /* Just call a subset of vital functions in case we are re-entering
      * the event loop from processEventsWhileBlocked(). Note that in this
      * case we keep track of the number of events we are processing, since
@@ -4067,12 +4069,6 @@ void initServer(void) {
         }
     }
 
-    /* We have to initialize storage providers after the cluster has been initialized */
-    for (int idb = 0; idb < cserver.dbnum; ++idb)
-    {
-        g_pserver->db[idb]->storageProviderInitialize();
-    }
-
     saveMasterStatusToStorage(false); // eliminate the repl-offset field
     
     /* Initialize ACL default password if it exists */
@@ -4085,6 +4081,15 @@ void initServer(void) {
  * Thread Local Storage initialization collides with dlopen call.
  * see: https://sourceware.org/bugzilla/show_bug.cgi?id=19329 */
 void InitServerLast() {
+
+    /* We have to initialize storage providers after the cluster has been initialized */
+    moduleFireServerEvent(REDISMODULE_EVENT_LOADING, REDISMODULE_SUBEVENT_LOADING_FLASH_START, NULL);
+    for (int idb = 0; idb < cserver.dbnum; ++idb)
+    {
+        g_pserver->db[idb]->storageProviderInitialize();
+    }
+    moduleFireServerEvent(REDISMODULE_EVENT_LOADING, REDISMODULE_SUBEVENT_LOADING_ENDED, NULL);
+
     bioInit();
     set_jemalloc_bg_thread(cserver.jemalloc_bg_thread);
     g_pserver->initial_memory_usage = zmalloc_used_memory();
@@ -5167,7 +5172,7 @@ int prepareForShutdown(int flags) {
          * to unlink file actully) in background thread.
          * The temp rdb file fd may won't be closed when redis exits quickly,
          * but OS will close this fd when process exits. */
-        rdbRemoveTempFile(g_pserver->child_pid, 0);
+        rdbRemoveTempFile(g_pserver->rdbThreadVars.tmpfileNum, 0);
     }
 
     /* Kill module child if there is one. */
@@ -5467,30 +5472,30 @@ NULL
 
 /* Convert an amount of bytes into a human readable string in the form
  * of 100B, 2G, 100M, 4K, and so forth. */
-void bytesToHuman(char *s, unsigned long long n) {
+void bytesToHuman(char *s, unsigned long long n, size_t bufsize) {
     double d;
 
     if (n < 1024) {
         /* Bytes */
-        snprintf(s,256,"%lluB",n);
+        snprintf(s,bufsize,"%lluB",n);
     } else if (n < (1024*1024)) {
         d = (double)n/(1024);
-        snprintf(s,256,"%.2fK",d);
+        snprintf(s,bufsize,"%.2fK",d);
     } else if (n < (1024LL*1024*1024)) {
         d = (double)n/(1024*1024);
-        snprintf(s,256,"%.2fM",d);
+        snprintf(s,bufsize,"%.2fM",d);
     } else if (n < (1024LL*1024*1024*1024)) {
         d = (double)n/(1024LL*1024*1024);
-        snprintf(s,256,"%.2fG",d);
+        snprintf(s,bufsize,"%.2fG",d);
     } else if (n < (1024LL*1024*1024*1024*1024)) {
         d = (double)n/(1024LL*1024*1024*1024);
-        snprintf(s,256,"%.2fT",d);
+        snprintf(s,bufsize,"%.2fT",d);
     } else if (n < (1024LL*1024*1024*1024*1024*1024)) {
         d = (double)n/(1024LL*1024*1024*1024*1024);
-        snprintf(s,256,"%.2fP",d);
+        snprintf(s,bufsize,"%.2fP",d);
     } else {
         /* Let's hope we never need this */
-        snprintf(s,256,"%lluB",n);
+        snprintf(s,bufsize,"%lluB",n);
     }
 }
 
@@ -5666,13 +5671,13 @@ sds genRedisInfoString(const char *section) {
         if (zmalloc_used > g_pserver->stat_peak_memory)
             g_pserver->stat_peak_memory = zmalloc_used;
 
-        bytesToHuman(hmem,zmalloc_used);
-        bytesToHuman(peak_hmem,g_pserver->stat_peak_memory);
-        bytesToHuman(total_system_hmem,total_system_mem);
-        bytesToHuman(used_memory_lua_hmem,memory_lua);
-        bytesToHuman(used_memory_scripts_hmem,mh->lua_caches);
-        bytesToHuman(used_memory_rss_hmem,g_pserver->cron_malloc_stats.process_rss);
-        bytesToHuman(maxmemory_hmem,g_pserver->maxmemory);
+        bytesToHuman(hmem,zmalloc_used,sizeof(hmem));
+        bytesToHuman(peak_hmem,g_pserver->stat_peak_memory,sizeof(peak_hmem));
+        bytesToHuman(total_system_hmem,total_system_mem,sizeof(total_system_hmem));
+        bytesToHuman(used_memory_lua_hmem,memory_lua,sizeof(used_memory_lua_hmem));
+        bytesToHuman(used_memory_scripts_hmem,mh->lua_caches,sizeof(used_memory_scripts_hmem));
+        bytesToHuman(used_memory_rss_hmem,g_pserver->cron_malloc_stats.process_rss,sizeof(used_memory_rss_hmem));
+        bytesToHuman(maxmemory_hmem,g_pserver->maxmemory,sizeof(maxmemory_hmem));
 
         if (sections++) info = sdscat(info,"\r\n");
         info = sdscatprintf(info,
@@ -6559,7 +6564,8 @@ void usage(void) {
 
 void redisAsciiArt(void) {
 #include "asciilogo.h"
-    char *buf = (char*)zmalloc(1024*16, MALLOC_LOCAL);
+    size_t bufsize = 1024*16;
+    char *buf = (char*)zmalloc(bufsize, MALLOC_LOCAL);
     const char *mode;
 
     if (g_pserver->cluster_enabled) mode = "cluster";
@@ -6581,7 +6587,7 @@ void redisAsciiArt(void) {
         );
     } else {
         sds motd = fetchMOTD(true, cserver.enable_motd);
-        snprintf(buf,1024*16,ascii_logo,
+        snprintf(buf,bufsize,ascii_logo,
             KEYDB_REAL_VERSION,
             redisGitSHA1(),
             strtol(redisGitDirty(),NULL,10) > 0,
@@ -6884,6 +6890,7 @@ int redisFork(int purpose) {
     latencyAddSampleIfNeeded("fork-lock",(ustime()-startWriteLock)/1000);
     if ((childpid = fork()) == 0) {
         /* Child */
+        aeForkLockInChild();
         aeReleaseForkLock();
         g_pserver->in_fork_child = purpose;
         setOOMScoreAdj(CONFIG_OOM_BGCHILD);

src/server.h

@@ -1204,12 +1204,13 @@ public:
     void disableKeyCache();
     bool keycacheIsEnabled();
 
-    bool prefetchKeysAsync(client *c, struct parsed_command &command, bool fExecOK);
+    void prefetchKeysAsync(client *c, struct parsed_command &command);
 
     bool FSnapshot() const { return m_spdbSnapshotHOLDER != nullptr; }
 
     std::unique_ptr<const StorageCache> CloneStorageCache() { return std::unique_ptr<const StorageCache>(m_spstorage->clone()); }
-    void bulkStorageInsert(char **rgKeys, size_t *rgcbKeys, char **rgVals, size_t *rgcbVals, size_t celem);
+    std::shared_ptr<StorageCache> getStorageCache() { return m_spstorage; }
+    void bulkDirectStorageInsert(char **rgKeys, size_t *rgcbKeys, char **rgVals, size_t *rgcbVals, size_t celem);
 
     dict_iter find_cached_threadsafe(const char *key) const;
 
@@ -1370,7 +1371,8 @@ struct redisDb : public redisDbPersistentDataSnapshot
     using redisDbPersistentData::FRehashing;
     using redisDbPersistentData::FTrackingChanges;
     using redisDbPersistentData::CloneStorageCache;
-    using redisDbPersistentData::bulkStorageInsert;
+    using redisDbPersistentData::bulkDirectStorageInsert;
+    using redisDbPersistentData::getStorageCache;
 
 public:
     const redisDbPersistentDataSnapshot *createSnapshot(uint64_t mvccCheckpoint, bool fOptional) {
@@ -2470,7 +2472,8 @@ struct redisServer {
     time_t lastbgsave_try;          /* Unix time of last attempted bgsave */
     time_t rdb_save_time_last;      /* Time used by last RDB save run. */
     time_t rdb_save_time_start;     /* Current RDB save start time. */
-    pid_t rdb_child_pid = -1;            /* Used only during fork bgsave */
+    mstime_t rdb_save_latency;      /* Used to track end to end latency of rdb save*/
+    pid_t rdb_child_pid = -1;       /* Used only during fork bgsave */
     int rdb_bgsave_scheduled;       /* BGSAVE when possible if true. */
     int rdb_child_type;             /* Type of save by active child. */
     int lastbgsave_status;          /* C_OK or C_ERR */
@@ -2728,6 +2731,8 @@ struct redisServer {
     long long repl_batch_offStart = -1;
     long long repl_batch_idxStart = -1;
 
+    long long rand_total_threshold;
+
     int config_soft_shutdown = false;
     bool soft_shutdown = false;
 
@@ -3254,6 +3259,7 @@ void receiveChildInfo(void);
 void executeWithoutGlobalLock(std::function<void()> func);
 int redisFork(int type);
 int hasActiveChildProcess();
+int hasActiveChildProcessOrBGSave();
 void resetChildState();
 int isMutuallyExclusiveChildType(int type);
 
@@ -3892,13 +3898,13 @@ void incrementMvccTstamp();
 
 #if __GNUC__ >= 7 && !defined(NO_DEPRECATE_FREE)
  [[deprecated]]
-void *calloc(size_t count, size_t size);
+void *calloc(size_t count, size_t size) noexcept;
  [[deprecated]]
-void free(void *ptr);
+void free(void *ptr) noexcept;
  [[deprecated]]
-void *malloc(size_t size);
+void *malloc(size_t size) noexcept;
  [[deprecated]]
-void *realloc(void *ptr, size_t size);
+void *realloc(void *ptr, size_t size) noexcept;
 #endif
 
 /* Debugging stuff */

src/storage/rocksdb.cpp

@@ -168,6 +168,28 @@ bool RocksDBStorageProvider::enumerate(callback fn) const
     return !it->Valid();
 }
 
+size_t RocksDBStorageProvider::stateful_enumerate(callback fn)
+{
+    if (m_iter == nullptr || !m_iter->Valid()) {
+        m_iter = std::unique_ptr<rocksdb::Iterator>(m_spdb->NewIterator(ReadOptions(), m_spcolfamily.get()));
+        m_iter->SeekToFirst();
+    }
+    size_t count = 0;
+    while (m_iter->Valid()) {
+        if (FInternalKey(m_iter->key().data(), m_iter->key().size()))
+            continue;
+
+        if (fn(m_iter->key().data(), m_iter->key().size(), m_iter->value().data(), m_iter->value().size())) {
+            m_iter->Next();
+            ++count;
+        } else {
+            break;
+        }
+    }
+    assert(m_iter->status().ok()); // Check for any errors found during the scan
+    return count;
+}
+
 const IStorage *RocksDBStorageProvider::clone() const
 {
     std::unique_lock<fastlock> l(m_lock);

src/storage/rocksdb.h

@@ -22,6 +22,7 @@ class RocksDBStorageProvider : public IStorage
     rocksdb::ReadOptions m_readOptionsTemplate;
     size_t m_count = 0;
     mutable fastlock m_lock {"RocksDBStorageProvider"};
+    std::unique_ptr<rocksdb::Iterator> m_iter;
 
 public:
     RocksDBStorageProvider(RocksDBStorageFactory *pfactory, std::shared_ptr<rocksdb::DB> &spdb, std::shared_ptr<rocksdb::ColumnFamilyHandle> &spcolfam, const rocksdb::Snapshot *psnapshot, size_t count);
@@ -32,6 +33,7 @@ public:
     virtual void retrieve(const char *key, size_t cchKey, callbackSingle fn) const override;
     virtual size_t clear() override;
     virtual bool enumerate(callback fn) const override;
+    virtual size_t stateful_enumerate(callback fn) override;
 
     virtual const IStorage *clone() const override;
 

src/storage/teststorageprovider.cpp

@@ -73,6 +73,26 @@ bool TestStorageProvider::enumerate(callback fn) const
     }
     return fAll;
 }
+
+size_t TestStorageProvider::stateful_enumerate(callback fn)
+{
+    size_t count = 0;
+    if (m_iter == m_map.end())
+        m_iter = m_map.begin();
+
+    while (m_iter != m_map.end())
+    {
+        if (fn(m_iter->first.data(), m_iter->first.size(), m_iter->second.data(), m_iter->second.size()))
+        {
+            m_iter++;
+            count++;
+        } else
+        {
+            break;
+        }
+    }
+    return count;
+}
     
 size_t TestStorageProvider::count() const
 {

src/storage/teststorageprovider.h

@@ -14,6 +14,7 @@ class TestStorageFactory : public IStorageFactory
 class TestStorageProvider final : public IStorage
 {
     std::unordered_map<std::string, std::string> m_map;
+    std::unordered_map<std::string, std::string>::iterator m_iter = m_map.end();
 
 public:
     TestStorageProvider();
@@ -24,6 +25,7 @@ public:
     virtual void retrieve(const char *key, size_t cchKey, callbackSingle fn) const override;
     virtual size_t clear() override;
     virtual bool enumerate(callback fn) const override;
+    virtual size_t stateful_enumerate(callback fn) override;
     virtual size_t count() const override;
 
     virtual void flush() override;

src/t_hash.cpp

@@ -729,6 +729,10 @@ void hincrbyfloatCommand(client *c) {
     unsigned int vlen;
 
     if (getLongDoubleFromObjectOrReply(c,c->argv[3],&incr,NULL) != C_OK) return;
+    if (isnan(incr) || isinf(incr)) {
+        addReplyError(c,"value is NaN or Infinity");
+        return;
+    }
     if ((o = hashTypeLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
     if (hashTypeGetValue(o,szFromObj(c->argv[2]),&vstr,&vlen,&ll) == C_OK) {
         if (vstr) {
@@ -1062,6 +1066,8 @@ void hrandfieldWithCountCommand(client *c, long l, int withvalues) {
                 addReplyBulkCBuffer(c, key, sdslen(key));
                 if (withvalues)
                     addReplyBulkCBuffer(c, value, sdslen(value));
+                if (c->flags & CLIENT_CLOSE_ASAP)
+                    break;
             }
         } else if (hash->encoding == OBJ_ENCODING_ZIPLIST) {
             ziplistEntry *keys, *vals = NULL;
@@ -1075,6 +1081,8 @@ void hrandfieldWithCountCommand(client *c, long l, int withvalues) {
                 count -= sample_count;
                 ziplistRandomPairs((unsigned char*)ptrFromObj(hash), sample_count, keys, vals);
                 harndfieldReplyWithZiplist(c, sample_count, keys, vals);
+                if (c->flags & CLIENT_CLOSE_ASAP)
+                    break;
             }
             zfree(keys);
             zfree(vals);
@@ -1221,13 +1229,13 @@ void hrandfieldCommand(client *c) {
     ziplistEntry ele;
 
     if (c->argc >= 3) {
-        if (getLongFromObjectOrReply(c,c->argv[2],&l,NULL) != C_OK) return;
+        if (getRangeLongFromObjectOrReply(c,c->argv[2],-LONG_MAX,LONG_MAX,&l,NULL) != C_OK) return;
         if (c->argc > 4 || (c->argc == 4 && strcasecmp(szFromObj(c->argv[3]),"withvalues"))) {
             addReplyErrorObject(c,shared.syntaxerr);
             return;
         } else if (c->argc == 4) {
             withvalues = 1;
-            if (l < LONG_MIN/2 || l > LONG_MAX/2) {
+            if (l < -g_pserver->rand_total_threshold || l > g_pserver->rand_total_threshold) {
                 addReplyError(c,"value is out of range");
                 return;
             }

src/t_set.cpp

@@ -672,7 +672,11 @@ void srandmemberWithCountCommand(client *c) {
 
     dict *d;
 
-    if (getLongFromObjectOrReply(c,c->argv[2],&l,NULL) != C_OK) return;
+    if (getRangeLongFromObjectOrReply(c,c->argv[2],-LONG_MAX,LONG_MAX,&l,NULL) != C_OK) return;
+    if (l < -g_pserver->rand_total_threshold || l > g_pserver->rand_total_threshold) {
+        addReplyError(c,"value is out of range");
+        return;
+    }
     if (l >= 0) {
         count = (unsigned long) l;
     } else {
@@ -706,6 +710,8 @@ void srandmemberWithCountCommand(client *c) {
             } else {
                 addReplyBulkCBuffer(c,ele,sdslen(ele));
             }
+            if (c->flags & CLIENT_CLOSE_ASAP)
+                break;
         }
         return;
     }

src/t_zset.cpp

@@ -4054,6 +4054,8 @@ void zrandmemberWithCountCommand(client *c, long l, int withscores) {
                 addReplyBulkCBuffer(c, key, sdslen(key));
                 if (withscores)
                     addReplyDouble(c, *(double*)dictGetVal(de));
+                if (c->flags & CLIENT_CLOSE_ASAP)
+                    break;
             }
         } else if (zsetobj->encoding == OBJ_ENCODING_ZIPLIST) {
             ziplistEntry *keys, *vals = NULL;
@@ -4067,6 +4069,8 @@ void zrandmemberWithCountCommand(client *c, long l, int withscores) {
                 count -= sample_count;
                 ziplistRandomPairs((unsigned char*)ptrFromObj(zsetobj), sample_count, keys, vals);
                 zarndmemberReplyWithZiplist(c, sample_count, keys, vals);
+                if (c->flags & CLIENT_CLOSE_ASAP)
+                    break;
             }
             zfree(keys);
             zfree(vals);
@@ -4210,13 +4214,13 @@ void zrandmemberCommand(client *c) {
     ziplistEntry ele;
 
     if (c->argc >= 3) {
-        if (getLongFromObjectOrReply(c,c->argv[2],&l,NULL) != C_OK) return;
+        if (getRangeLongFromObjectOrReply(c,c->argv[2],-LONG_MAX,LONG_MAX,&l,NULL) != C_OK) return;
         if (c->argc > 4 || (c->argc == 4 && strcasecmp(szFromObj(c->argv[3]),"withscores"))) {
             addReplyErrorObject(c,shared.syntaxerr);
             return;
         } else if (c->argc == 4) {
             withscores = 1;
-            if (l < LONG_MIN/2 || l > LONG_MAX/2) {
+            if (l < -g_pserver->rand_total_threshold || l > g_pserver->rand_total_threshold) {
                 addReplyError(c,"value is out of range");
                 return;
             }

src/tls.cpp

@@ -616,8 +616,9 @@ bool tlsValidateCertificateName(tls_connection* conn){
     /* If neither the CN nor the SANs match, update the SSL error and return false */
     conn->c.last_errno = 0;
     if (conn->ssl_error) zfree(conn->ssl_error);
-    conn->ssl_error = (char*)zmalloc(512);
-    snprintf(conn->ssl_error, 512, "Client CN (%s) and SANs not found in allowlist.", commonName);
+    size_t bufsize = 512;
+    conn->ssl_error = (char*)zmalloc(bufsize);
+    snprintf(conn->ssl_error, bufsize, "Client CN (%s) and SANs not found in allowlist.", commonName);
     return false;
 }
 

src/ziplist.c

@@ -1704,17 +1704,17 @@ static unsigned char *createIntList() {
     unsigned char *zl = ziplistNew();
     char buf[32];
 
-    snprintf(buf, 32, "100");
+    snprintf(buf, sizeof(buf), "100");
     zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL);
-    snprintf(buf, 32, "128000");
+    snprintf(buf, sizeof(buf), "128000");
     zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL);
-    snprintf(buf, 32, "-100");
+    snprintf(buf, sizeof(buf), "-100");
     zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_HEAD);
-    snprintf(buf, 32, "4294967296");
+    snprintf(buf, sizeof(buf), "4294967296");
     zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_HEAD);
-    snprintf(buf, 32, "non integer");
+    snprintf(buf, sizeof(buf), "non integer");
     zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL);
-    snprintf(buf, 32, "much much longer non integer");
+    snprintf(buf, sizeof(buf), "much much longer non integer");
     zl = ziplistPush(zl, (unsigned char*)buf, strlen(buf), ZIPLIST_TAIL);
     return zl;
 }
@@ -2228,7 +2228,7 @@ int ziplistTest(int argc, char **argv, int accurate) {
         char buf[32];
         int i,len;
         for (i = 0; i < 1000; i++) {
-            len = snprintf(buf,32,"%d",i);
+            len = snprintf(buf,sizeof(buf),"%d",i);
             zl = ziplistPush(zl,(unsigned char*)buf,len,ZIPLIST_TAIL);
         }
         for (i = 0; i < 1000; i++) {
@@ -2375,13 +2375,13 @@ int ziplistTest(int argc, char **argv, int accurate) {
                 } else {
                     switch(rand() % 3) {
                     case 0:
-                        buflen = snprintf(buf,1024,"%lld",(0LL + rand()) >> 20);
+                        buflen = snprintf(buf,sizeof(buf),"%lld",(0LL + rand()) >> 20);
                         break;
                     case 1:
-                        buflen = snprintf(buf,1024,"%lld",(0LL + rand()));
+                        buflen = snprintf(buf,sizeof(buf),"%lld",(0LL + rand()));
                         break;
                     case 2:
-                        buflen = snprintf(buf,1024,"%lld",(0LL + rand()) << 20);
+                        buflen = snprintf(buf,sizeof(buf),"%lld",(0LL + rand()) << 20);
                         break;
                     default:
                         assert(NULL);
@@ -2410,7 +2410,7 @@ int ziplistTest(int argc, char **argv, int accurate) {
 
                 assert(ziplistGet(p,&sstr,&slen,&sval));
                 if (sstr == NULL) {
-                    buflen = snprintf(buf,1024,"%lld",sval);
+                    buflen = snprintf(buf,sizeof(buf),"%lld",sval);
                 } else {
                     buflen = slen;
                     memcpy(buf,sstr,buflen);

src/zmalloc.cpp

@@ -80,12 +80,12 @@ static_assert((PREFIX_SIZE % 16) == 0, "Our prefix must be modulo 16-bytes or ou
 #define realloc(ptr,size, type) tc_realloc(ptr,size)
 #define free(ptr) tc_free(ptr)
 #elif defined(USE_JEMALLOC)
-#define malloc(size, type) je_malloc(size)
-#define calloc(count,size,type) je_calloc(count,size)
-#define realloc(ptr,size,type) je_realloc(ptr,size)
-#define free(ptr) je_free(ptr)
-#define mallocx(size,flags) je_mallocx(size,flags)
-#define dallocx(ptr,flags) je_dallocx(ptr,flags)
+#define malloc(size, type) malloc(size)
+#define calloc(count,size,type) calloc(count,size)
+#define realloc(ptr,size,type) realloc(ptr,size)
+#define free(ptr) free(ptr)
+#define mallocx(size,flags) mallocx(size,flags)
+#define dallocx(ptr,flags) dallocx(ptr,flags)
 #else
 #define malloc(size, type) malloc(size)
 #define calloc(count,size,type) calloc(count,size)
@@ -370,7 +370,7 @@ size_t zmalloc_get_rss(void) {
     int fd, count;
     char *p, *x;
 
-    snprintf(filename,256,"/proc/%ld/stat",(long) getpid());
+    snprintf(filename,sizeof(filename),"/proc/%ld/stat",(long) getpid());
     if ((fd = open(filename,O_RDONLY)) == -1) return 0;
     if (read(fd,buf,4096) <= 0) {
         close(fd);
@@ -462,7 +462,7 @@ size_t zmalloc_get_rss(void) {
     char filename[256];
     int fd;
 
-    snprintf(filename,256,"/proc/%ld/psinfo",(long) getpid());
+    snprintf(filename,sizeof(filename),"/proc/%ld/psinfo",(long) getpid());
 
     if ((fd = open(filename,O_RDONLY)) == -1) return 0;
     if (ioctl(fd, PIOCPSINFO, &info) == -1) {
@@ -494,17 +494,17 @@ int zmalloc_get_allocator_info(size_t *allocated,
     *allocated = *resident = *active = 0;
     /* Update the statistics cached by mallctl. */
     sz = sizeof(epoch);
-    je_mallctl("epoch", &epoch, &sz, &epoch, sz);
+    mallctl("epoch", &epoch, &sz, &epoch, sz);
     sz = sizeof(size_t);
     /* Unlike RSS, this does not include RSS from shared libraries and other non
      * heap mappings. */
-    je_mallctl("stats.resident", resident, &sz, NULL, 0);
+    mallctl("stats.resident", resident, &sz, NULL, 0);
     /* Unlike resident, this doesn't not include the pages jemalloc reserves
      * for re-use (purge will clean that). */
-    je_mallctl("stats.active", active, &sz, NULL, 0);
+    mallctl("stats.active", active, &sz, NULL, 0);
     /* Unlike zmalloc_used_memory, this matches the stats.resident by taking
      * into account all allocations done by this process (not only zmalloc). */
-    je_mallctl("stats.allocated", allocated, &sz, NULL, 0);
+    mallctl("stats.allocated", allocated, &sz, NULL, 0);
     return 1;
 }
 
@@ -512,7 +512,7 @@ void set_jemalloc_bg_thread(int enable) {
     /* let jemalloc do purging asynchronously, required when there's no traffic 
      * after flushdb */
     char val = !!enable;
-    je_mallctl("background_thread", NULL, 0, &val, 1);
+    mallctl("background_thread", NULL, 0, &val, 1);
 }
 
 int jemalloc_purge() {
@@ -520,9 +520,9 @@ int jemalloc_purge() {
     char tmp[32];
     unsigned narenas = 0;
     size_t sz = sizeof(unsigned);
-    if (!je_mallctl("arenas.narenas", &narenas, &sz, NULL, 0)) {
-        snprintf(tmp, 32, "arena.%d.purge", narenas);
-        if (!je_mallctl(tmp, NULL, 0, NULL, 0))
+    if (!mallctl("arenas.narenas", &narenas, &sz, NULL, 0)) {
+        snprintf(tmp, sizeof(tmp), "arena.%d.purge", narenas);
+        if (!mallctl(tmp, NULL, 0, NULL, 0))
             return 0;
     }
     return -1;

src/zmalloc.h

@@ -55,7 +55,7 @@
 #include <jemalloc/jemalloc.h>
 #if (JEMALLOC_VERSION_MAJOR == 2 && JEMALLOC_VERSION_MINOR >= 1) || (JEMALLOC_VERSION_MAJOR > 2)
 #define HAVE_MALLOC_SIZE 1
-#define zmalloc_size(p) je_malloc_usable_size(p)
+#define zmalloc_size(p) malloc_usable_size(p)
 #else
 #error "Newer version of jemalloc required"
 #endif

tests/modules/Makefile

@@ -39,7 +39,8 @@ TEST_MODULES = \
     defragtest.so \
     hash.so \
     zset.so \
-    stream.so
+    stream.so \
+    load.so
 
 
 .PHONY: all

tests/modules/hooks.c

@@ -212,6 +212,7 @@ void loadingCallback(RedisModuleCtx *ctx, RedisModuleEvent e, uint64_t sub, void
         case REDISMODULE_SUBEVENT_LOADING_RDB_START: keyname = "loading-rdb-start"; break;
         case REDISMODULE_SUBEVENT_LOADING_AOF_START: keyname = "loading-aof-start"; break;
         case REDISMODULE_SUBEVENT_LOADING_REPL_START: keyname = "loading-repl-start"; break;
+        case REDISMODULE_SUBEVENT_LOADING_FLASH_START: keyname = "loading-flash-start"; break;
         case REDISMODULE_SUBEVENT_LOADING_ENDED: keyname = "loading-end"; break;
         case REDISMODULE_SUBEVENT_LOADING_FAILED: keyname = "loading-failed"; break;
     }

tests/modules/load.c

@@ -0,0 +1,94 @@
+/* Server hooks API example
+ *
+ * -----------------------------------------------------------------------------
+ *
+ * Copyright (c) 2019, Salvatore Sanfilippo <antirez at gmail dot com>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   * Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *   * Redistributions in binary form must reproduce the above copyright
+ *     notice, this list of conditions and the following disclaimer in the
+ *     documentation and/or other materials provided with the distribution.
+ *   * Neither the name of Redis nor the names of its contributors may be used
+ *     to endorse or promote products derived from this software without
+ *     specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#define REDISMODULE_EXPERIMENTAL_API
+#include "redismodule.h"
+
+size_t count, finalCount;
+
+/* Client state change callback. */
+void loadCallback(RedisModuleCtx *ctx, RedisModuleEvent e, uint64_t sub, void *data) {
+    REDISMODULE_NOT_USED(ctx);
+    REDISMODULE_NOT_USED(e);
+    REDISMODULE_NOT_USED(data);
+
+    if (sub == REDISMODULE_SUBEVENT_LOADING_FLASH_START || sub == REDISMODULE_SUBEVENT_LOADING_RDB_START || sub == REDISMODULE_SUBEVENT_LOADING_AOF_START || sub == REDISMODULE_SUBEVENT_LOADING_REPL_START) {
+        count = 0;
+        finalCount = 0;
+    } else if (sub == REDISMODULE_SUBEVENT_LOADING_ENDED) {
+        finalCount = count;
+    }
+}
+
+int loadKeyCallback(RedisModuleCtx *ctx, int type,  const char *event, RedisModuleString *key) {
+    REDISMODULE_NOT_USED(ctx);
+    REDISMODULE_NOT_USED(type);
+    REDISMODULE_NOT_USED(event);
+
+    const char *keyname = RedisModule_StringPtrLen(key, NULL);
+
+    RedisModule_Log(ctx, REDISMODULE_LOGLEVEL_NOTICE, "Loaded key: %s", keyname);
+
+    count++;
+    return 0;
+}
+
+int LoadCount_RedisCommand(RedisModuleCtx *ctx, RedisModuleString **argv, int argc) {
+    REDISMODULE_NOT_USED(argv);
+    RedisModule_AutoMemory(ctx); /* Use automatic memory management. */
+
+    if (argc != 1) return RedisModule_WrongArity(ctx);
+
+    RedisModule_ReplyWithLongLong(ctx, finalCount);
+
+    return REDISMODULE_OK;
+}
+
+/* This function must be present on each Redis module. It is used in order to
+ * register the commands into the Redis server. */
+int RedisModule_OnLoad(RedisModuleCtx *ctx, RedisModuleString **argv, int argc) {
+    REDISMODULE_NOT_USED(argv);
+    REDISMODULE_NOT_USED(argc);
+
+    if (RedisModule_Init(ctx,"load",1,REDISMODULE_APIVER_1)
+        == REDISMODULE_ERR) return REDISMODULE_ERR;
+
+    RedisModule_SubscribeToServerEvent(ctx,
+        RedisModuleEvent_Loading, loadCallback);
+    RedisModule_SubscribeToKeyspaceEvents(ctx,
+        REDISMODULE_NOTIFY_LOADED, loadKeyCallback);
+
+    if (RedisModule_CreateCommand(ctx, "load.count",
+        LoadCount_RedisCommand,"readonly",1,1,1) == REDISMODULE_ERR)
+        return REDISMODULE_ERR;
+    return REDISMODULE_OK;
+}

tests/unit/flash.tcl

@@ -120,6 +120,21 @@ if {$::flash_enabled} {
             assert_equal {2} [r scard set1]
         }
 
+        test { Keys in storage are expired } {
+            r flushdb
+            r psetex key1 10000 a
+            r psetex key2 10000 a
+            r psetex key3 10000 a
+            r flushall cache
+
+            after 1000
+            wait_for_condition 1000 50 {
+                [r dbsize] == 0
+            } else {
+                fail "keys in ssd only not expired"
+            }
+        }
+
         r flushall
         # If a weak storage memory model is set, wait for any pending snapshot writes to finish
         after 500 

tests/unit/introspection.tcl

@@ -162,15 +162,16 @@ start_server {tags {"introspection"}} {
             aof_rewrite_cpulist
             time-thread-priority
             bgsave_cpulist
-	        storage-cache-mode
-	        storage-provider-options
-	        use-fork
+	    storage-cache-mode
+	    storage-provider-options
+	    use-fork
             multi-master
             active-replica
             bind
             set-proc-title
             repl-backlog-disk-reserve
 	    tls-allowlist
+            db-s3-object
         }
 
         if {!$::tls} {

tests/unit/moduleapi/load.tcl

@@ -0,0 +1,15 @@
+set testmodule [file normalize tests/modules/load.so]
+
+if {$::flash_enabled} {
+    start_server [list tags [list "modules"] overrides [list storage-provider {flash ./rocks.db.master} databases 256 loadmodule $testmodule]] {
+        test "Module is notified of keys loaded from flash" {
+            r flushall
+            r set foo bar
+            r set bar foo
+            r set foobar barfoo
+            assert_equal [r load.count] 0
+            r debug reload
+            assert_equal [r load.count] 3
+        }
+    }
+}

tests/unit/obuf-limits.tcl

@@ -182,4 +182,12 @@ start_server {tags {"obuf-limits"} overrides { server-threads 1 }} {
         assert_equal "v2" [r get k2]
         assert_equal "v3" [r get k3]
     }
+
+    test "Obuf limit, HRANDFIELD with huge count stopped mid-run" {
+        r config set client-output-buffer-limit {normal 1000000 0 0}
+        r hset myhash a b
+        catch {r hrandfield myhash -999999999} e
+        assert_match "*I/O error*" $e
+        reconnect
+    }
 }

tests/unit/scripting.tcl

@@ -43,6 +43,15 @@ start_server {tags {"scripting"}} {
         r eval {return redis.call('get',KEYS[1])} 1 mykey
     } {myval}
 
+    test {EVAL - keys command works? } {
+        r eval {return redis.call('keys', 'test')} 0
+    }
+
+    test {EVAL - KeyDB global works } {
+        r eval {return keydb.call('get', KEYS[1])} 1 mykey
+        assert_equal [r eval {return redis.call('get',KEYS[1])} 1 mykey] [r eval {return keydb.call('get', KEYS[1])} 1 mykey]
+    }
+
     test {EVALSHA - Can we call a SHA1 if already defined?} {
         r evalsha fd758d1589d044dd850a6f05d52f2eefd27f033f 1 mykey
     } {myval}

tests/unit/type/hash.tcl

@@ -71,6 +71,8 @@ start_server {tags {"hash"}} {
     test "HRANDFIELD count overflow" {
         r hmset myhash a 1
         assert_error {*value is out of range*} {r hrandfield myhash -9223372036854770000 withvalues}
+        assert_error {*value is out of range*} {r hrandfield myhash -9223372036854775808 withvalues}
+        assert_error {*value is out of range*} {r hrandfield myhash -9223372036854775808}
     } {}
 
     test "HRANDFIELD with <count> against non existing key" {
@@ -831,4 +833,8 @@ start_server {tags {"hash"}} {
         set _ $k
     } {ZIP_INT_8B 127 ZIP_INT_16B 32767 ZIP_INT_32B 2147483647 ZIP_INT_64B 9223372036854775808 ZIP_INT_IMM_MIN 0 ZIP_INT_IMM_MAX 12}
 
+    test {HINCRBYFLOAT does not allow NaN or Infinity} {
+        assert_error "*value is NaN or Infinity*" {r hincrbyfloat hfoo field +inf}
+        assert_equal 0 [r exists hfoo]
+    }
 }

tests/unit/type/set.tcl

@@ -588,6 +588,11 @@ start_server {
         r srandmember nonexisting_key 100
     } {}
 
+    test "SRANDMEMBER count overflow" {
+        r sadd myset a
+        assert_error {*value is out of range*} {r srandmember myset -9223372036854775808}
+    } {}
+
     # Make sure we can distinguish between an empty array and a null response
     r readraw 1
 

tests/unit/type/zset.tcl

@@ -1717,6 +1717,8 @@ start_server {tags {"zset"}} {
     test "ZRANDMEMBER count overflow" {
         r zadd myzset 0 a
         assert_error {*value is out of range*} {r zrandmember myzset -9223372036854770000 withscores}
+        assert_error {*value is out of range*} {r zrandmember myzset -9223372036854775808 withscores}
+        assert_error {*value is out of range*} {r zrandmember myzset -9223372036854775808}
     } {}
 
     # Make sure we can distinguish between an empty array and a null response

utils/create-cluster/create-cluster

@@ -26,7 +26,11 @@ then
     while [ $((PORT < ENDPORT)) != "0" ]; do
         PORT=$((PORT+1))
         echo "Starting $PORT"
-        $BIN_PATH/keydb-server --port $PORT  --protected-mode $PROTECTED_MODE --cluster-enabled yes --cluster-config-file nodes-${PORT}.conf --cluster-node-timeout $TIMEOUT --appendonly yes --appendfilename appendonly-${PORT}.aof --dbfilename dump-${PORT}.rdb --logfile ${PORT}.log --daemonize yes ${ADDITIONAL_OPTIONS}
+        if [ "$2" == "flash" ]
+        then
+            ADDITIONAL_OPTIONS="--save \"\" \"\" \"\" --semi-ordered-set-bucket-size 8 --client-output-buffer-limit replica 1 1 0 --maxmemory 100000000 --storage-provider flash ./$PORT.flash"
+        fi
+        $BIN_PATH/keydb-server --server-threads 4 --port $PORT  --protected-mode $PROTECTED_MODE --cluster-enabled yes --cluster-config-file nodes-${PORT}.conf --cluster-node-timeout $TIMEOUT --dbfilename dump-${PORT}.rdb --logfile ${PORT}.log --daemonize yes ${ADDITIONAL_OPTIONS}
     done
     exit 0
 fi
@@ -97,6 +101,8 @@ then
     rm -rf appendonly*.aof
     rm -rf dump*.rdb
     rm -rf nodes*.conf
+    rm -rf *.flash
+    rm -rf temp*.rdb
     exit 0
 fi